diff --git a/accounts/abi/bind/backends/simulated.go b/accounts/abi/bind/backends/simulated.go
index 146b209f7..60e39a34e 100644
--- a/accounts/abi/bind/backends/simulated.go
+++ b/accounts/abi/bind/backends/simulated.go
@@ -181,7 +181,6 @@ func (b *SimulatedBackend) CodeAt(ctx context.Context, contract common.Address,
 	if err != nil {
 		return nil, err
 	}
-
 	return stateDB.GetCode(contract), nil
 }
 
@@ -194,7 +193,6 @@ func (b *SimulatedBackend) BalanceAt(ctx context.Context, contract common.Addres
 	if err != nil {
 		return nil, err
 	}
-
 	return stateDB.GetBalance(contract), nil
 }
 
@@ -207,7 +205,6 @@ func (b *SimulatedBackend) NonceAt(ctx context.Context, contract common.Address,
 	if err != nil {
 		return 0, err
 	}
-
 	return stateDB.GetNonce(contract), nil
 }
 
@@ -220,7 +217,6 @@ func (b *SimulatedBackend) StorageAt(ctx context.Context, contract common.Addres
 	if err != nil {
 		return nil, err
 	}
-
 	val := stateDB.GetState(contract, key)
 	return val[:], nil
 }
@@ -667,7 +663,10 @@ func (b *SimulatedBackend) SendTransaction(ctx context.Context, tx *types.Transa
 		}
 		block.AddTxWithChain(b.blockchain, tx)
 	}, true)
-	stateDB, _ := b.blockchain.State()
+	stateDB, err := b.blockchain.State()
+	if err != nil {
+		return err
+	}
 
 	b.pendingBlock = blocks[0]
 	b.pendingState, _ = state.New(b.pendingBlock.Root(), stateDB.Database(), nil)
@@ -782,11 +781,13 @@ func (b *SimulatedBackend) AdjustTime(adjustment time.Duration) error {
 	blocks, _ := core.GenerateChain(b.config, b.blockchain.CurrentBlock(), ethash.NewFaker(), b.database, 1, func(number int, block *core.BlockGen) {
 		block.OffsetTime(int64(adjustment.Seconds()))
 	}, true)
-	stateDB, _ := b.blockchain.State()
+	stateDB, err := b.blockchain.State()
+	if err != nil {
+		return err
+	}
 
 	b.pendingBlock = blocks[0]
 	b.pendingState, _ = state.New(b.pendingBlock.Root(), stateDB.Database(), nil)
-
 	return nil
 }
 
diff --git a/cmd/devp2p/internal/ethtest/suite_test.go b/cmd/devp2p/internal/ethtest/suite_test.go
index 55e386c4a..4754e81e1 100644
--- a/cmd/devp2p/internal/ethtest/suite_test.go
+++ b/cmd/devp2p/internal/ethtest/suite_test.go
@@ -99,6 +99,7 @@ func setupGeth(stack *node.Node) error {
 	if err != nil {
 		return err
 	}
+	backend.SetSynced()
 
 	_, err = backend.BlockChain().InsertChain(chain.blocks[1:], nil)
 	return err
diff --git a/core/blockchain.go b/core/blockchain.go
index d1ec0cedc..5881713f6 100644
--- a/core/blockchain.go
+++ b/core/blockchain.go
@@ -374,12 +374,11 @@ func NewBlockChain(db ethdb.Database, cacheConfig *CacheConfig, genesis *Genesis
 	if !bc.HasState(head.Root()) {
 		if head.NumberU64() == 0 {
 			// The genesis state is missing, which is only possible in the path-based
-			// scheme. This situation occurs when the state syncer overwrites it.
-			//
-			// The solution is to reset the state to the genesis state. Although it may not
-			// match the sync target, the state healer will later address and correct any
-			// inconsistencies.
-			bc.resetState()
+			// scheme. This situation occurs when the initial state sync is not finished
+			// yet, or the chain head is rewound below the pivot point. In both scenario,
+			// there is no possible recovery approach except for rerunning a snap sync.
+			// Do nothing here until the state syncer picks it up.
+			log.Info("Genesis state is missing, wait state sync")
 		} else {
 			// Head state is missing, before the state recovery, find out the
 			// disk layer point of snapshot(if it's enabled). Make sure the
@@ -675,28 +674,6 @@ func (bc *BlockChain) SetHead(head uint64) error {
 	return err
 }
 
-// resetState resets the persistent state to genesis state if it's not present.
-func (bc *BlockChain) resetState() {
-	// Short circuit if the genesis state is already present.
-	root := bc.genesisBlock.Root()
-	if bc.HasState(root) {
-		return
-	}
-	// Reset the state database to empty for committing genesis state.
-	// Note, it should only happen in path-based scheme and Reset function
-	// is also only call-able in this mode.
-	if bc.triedb.Scheme() == rawdb.PathScheme {
-		if err := bc.triedb.Reset(types.EmptyRootHash); err != nil {
-			log.Crit("Failed to clean state", "err", err) // Shouldn't happen
-		}
-	}
-	// Write genesis state into database.
-	if err := CommitGenesisState(bc.db, bc.triedb, bc.genesisBlock.Hash()); err != nil {
-		log.Crit("Failed to commit genesis state", "err", err)
-	}
-	log.Info("Reset state to genesis", "root", root)
-}
-
 // setHeadBeyondRoot rewinds the local chain to a new head with the extra condition
 // that the rewind must pass the specified state root. This method is meant to be
 // used when rewinding with snapshots enabled to ensure that we go back further than
@@ -728,7 +705,6 @@ func (bc *BlockChain) setHeadBeyondRoot(head uint64, root common.Hash, repair bo
 			if newHeadBlock == nil {
 				log.Error("Gap in the chain, rewinding to genesis", "number", header.Number, "hash", header.Hash())
 				newHeadBlock = bc.genesisBlock
-				bc.resetState()
 			} else {
 				// Block exists, keep rewinding until we find one with state,
 				// keeping rewinding until we exceed the optional threshold
@@ -758,16 +734,14 @@ func (bc *BlockChain) setHeadBeyondRoot(head uint64, root common.Hash, repair bo
 						}
 					}
 					if beyondRoot || newHeadBlock.NumberU64() == 0 {
-						if newHeadBlock.NumberU64() == 0 {
-							bc.resetState()
-						} else if !bc.HasState(newHeadBlock.Root()) {
+						if !bc.HasState(newHeadBlock.Root()) && bc.stateRecoverable(newHeadBlock.Root()) {
 							// Rewind to a block with recoverable state. If the state is
 							// missing, run the state recovery here.
 							if err := bc.triedb.Recover(newHeadBlock.Root()); err != nil {
 								log.Crit("Failed to rollback state", "err", err) // Shouldn't happen
 							}
+							log.Debug("Rewound to block with state", "number", newHeadBlock.NumberU64(), "hash", newHeadBlock.Hash())
 						}
-						log.Debug("Rewound to block with state", "number", newHeadBlock.NumberU64(), "hash", newHeadBlock.Hash())
 						break
 					}
 					log.Debug("Skipping block with threshold state", "number", newHeadBlock.NumberU64(), "hash", newHeadBlock.Hash(), "root", newHeadBlock.Root())
@@ -782,6 +756,15 @@ func (bc *BlockChain) setHeadBeyondRoot(head uint64, root common.Hash, repair bo
 			// to low, so it's safe the update in-memory markers directly.
 			bc.currentBlock.Store(newHeadBlock)
 			headBlockGauge.Update(int64(newHeadBlock.NumberU64()))
+
+			// The head state is missing, which is only possible in the path-based
+			// scheme. This situation occurs when the chain head is rewound below
+			// the pivot point. In this scenario, there is no possible recovery
+			// approach except for rerunning a snap sync. Do nothing here until the
+			// state syncer picks it up.
+			if !bc.HasState(newHeadBlock.Root()) {
+				log.Info("Chain is stateless, wait state sync", "number", newHeadBlock.Number(), "hash", newHeadBlock.Hash())
+			}
 		}
 		// Rewind the fast block in a simpleton way to the target head
 		if currentFastBlock := bc.CurrentFastBlock(); currentFastBlock != nil && header.Number.Uint64() < currentFastBlock.NumberU64() {
@@ -865,7 +848,7 @@ func (bc *BlockChain) FastSyncCommitHead(hash common.Hash) error {
 	// Reset the trie database with the fresh fast synced state.
 	root := block.Root()
 	if bc.triedb.Scheme() == rawdb.PathScheme {
-		if err := bc.triedb.Reset(root); err != nil {
+		if err := bc.triedb.Enable(root); err != nil {
 			return err
 		}
 	}
diff --git a/core/rawdb/accessors_sync.go b/core/rawdb/accessors_sync.go
new file mode 100644
index 000000000..54e5967d8
--- /dev/null
+++ b/core/rawdb/accessors_sync.go
@@ -0,0 +1,44 @@
+// Copyright 2022 The go-ethereum Authors
+// This file is part of the go-ethereum library.
+//
+// The go-ethereum library is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Lesser General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// The go-ethereum library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Lesser General Public License for more details.
+//
+// You should have received a copy of the GNU Lesser General Public License
+// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
+
+package rawdb
+
+import (
+	"github.com/ethereum/go-ethereum/ethdb"
+	"github.com/ethereum/go-ethereum/log"
+)
+
+const (
+	StateSyncUnknown  = uint8(0) // flags the state snap sync is unknown
+	StateSyncRunning  = uint8(1) // flags the state snap sync is not completed yet
+	StateSyncFinished = uint8(2) // flags the state snap sync is completed
+)
+
+// ReadSnapSyncStatusFlag retrieves the state snap sync status flag.
+func ReadSnapSyncStatusFlag(db ethdb.KeyValueReader) uint8 {
+	blob, err := db.Get(snapSyncStatusFlagKey)
+	if err != nil || len(blob) != 1 {
+		return StateSyncUnknown
+	}
+	return blob[0]
+}
+
+// WriteSnapSyncStatusFlag stores the state snap sync status flag into database.
+func WriteSnapSyncStatusFlag(db ethdb.KeyValueWriter, flag uint8) {
+	if err := db.Put(snapSyncStatusFlagKey, []byte{flag}); err != nil {
+		log.Crit("Failed to store sync status flag", "err", err)
+	}
+}
diff --git a/core/rawdb/accessors_trie.go b/core/rawdb/accessors_trie.go
index 470615660..14b3a96ba 100644
--- a/core/rawdb/accessors_trie.go
+++ b/core/rawdb/accessors_trie.go
@@ -88,6 +88,16 @@ func HasAccountTrieNode(db ethdb.KeyValueReader, path []byte, hash common.Hash)
 	return h.hash(data) == hash
 }
 
+// ExistsAccountTrieNode checks the presence of the account trie node with the
+// specified node path, regardless of the node hash.
+func ExistsAccountTrieNode(db ethdb.KeyValueReader, path []byte) bool {
+	has, err := db.Has(accountTrieNodeKey(path))
+	if err != nil {
+		return false
+	}
+	return has
+}
+
 // WriteAccountTrieNode writes the provided account trie node into database.
 func WriteAccountTrieNode(db ethdb.KeyValueWriter, path []byte, node []byte) {
 	if err := db.Put(accountTrieNodeKey(path), node); err != nil {
@@ -126,6 +136,16 @@ func HasStorageTrieNode(db ethdb.KeyValueReader, accountHash common.Hash, path [
 	return h.hash(data) == hash
 }
 
+// ExistsStorageTrieNode checks the presence of the storage trie node with the
+// specified account hash and node path, regardless of the node hash.
+func ExistsStorageTrieNode(db ethdb.KeyValueReader, accountHash common.Hash, path []byte) bool {
+	has, err := db.Has(storageTrieNodeKey(accountHash, path))
+	if err != nil {
+		return false
+	}
+	return has
+}
+
 // WriteStorageTrieNode writes the provided storage trie node into database.
 func WriteStorageTrieNode(db ethdb.KeyValueWriter, accountHash common.Hash, path []byte, node []byte) {
 	if err := db.Put(storageTrieNodeKey(accountHash, path), node); err != nil {
diff --git a/core/rawdb/database.go b/core/rawdb/database.go
index 522fa33ca..9ad0287d5 100644
--- a/core/rawdb/database.go
+++ b/core/rawdb/database.go
@@ -535,7 +535,7 @@ func InspectDatabase(db ethdb.Database, keyPrefix, keyStart []byte) error {
 				fastTrieProgressKey, snapshotDisabledKey, SnapshotRootKey, snapshotJournalKey,
 				snapshotGeneratorKey, snapshotRecoveryKey, txIndexTailKey, fastTxLookupLimitKey,
 				uncleanShutdownKey, badBlockKey, highestFinalityVoteKey, storeInternalTxsEnabledKey,
-				snapshotSyncStatusKey, persistentStateIDKey, trieJournalKey, snapshotSyncStatusKey,
+				snapshotSyncStatusKey, persistentStateIDKey, trieJournalKey, snapshotSyncStatusKey, snapSyncStatusFlagKey,
 			} {
 				if bytes.Equal(key, meta) {
 					metadata.Add(size)
diff --git a/core/rawdb/schema.go b/core/rawdb/schema.go
index 84b2bc8e5..3c99d1db6 100644
--- a/core/rawdb/schema.go
+++ b/core/rawdb/schema.go
@@ -90,6 +90,9 @@ var (
 
 	snapshotConsortiumPrefix = []byte("consortium-") // key = ConsortiumSnapshotPrefix + block hash
 
+	// snapSyncStatusFlagKey flags that status of snap sync.
+	snapSyncStatusFlagKey = []byte("SnapSyncStatus")
+
 	// Data item prefixes (use single byte to avoid mixing data types, avoid `i`, used for indexes).
 	headerPrefix       = []byte("h") // headerPrefix + num (uint64 big endian) + hash -> header
 	headerTDSuffix     = []byte("t") // headerPrefix + num (uint64 big endian) + hash + headerTDSuffix -> td
diff --git a/eth/api_backend.go b/eth/api_backend.go
index b7e4b1c2a..60a1869e7 100644
--- a/eth/api_backend.go
+++ b/eth/api_backend.go
@@ -168,7 +168,10 @@ func (b *EthAPIBackend) StateAndHeaderByNumber(ctx context.Context, number rpc.B
 		return nil, nil, errors.New("header not found")
 	}
 	stateDb, err := b.eth.BlockChain().StateAt(header.Root)
-	return stateDb, header, err
+	if err != nil {
+		return nil, nil, err
+	}
+	return stateDb, header, nil
 }
 
 func (b *EthAPIBackend) StateAndHeaderByNumberOrHash(ctx context.Context, blockNrOrHash rpc.BlockNumberOrHash) (*state.StateDB, *types.Header, error) {
@@ -187,7 +190,10 @@ func (b *EthAPIBackend) StateAndHeaderByNumberOrHash(ctx context.Context, blockN
 			return nil, nil, errors.New("hash is not currently canonical")
 		}
 		stateDb, err := b.eth.BlockChain().StateAt(header.Root)
-		return stateDb, header, err
+		if err != nil {
+			return nil, nil, err
+		}
+		return stateDb, header, nil
 	}
 	return nil, nil, errors.New("invalid arguments; neither block nor hash specified")
 }
diff --git a/eth/backend.go b/eth/backend.go
index 6345ad809..cc6e36b33 100644
--- a/eth/backend.go
+++ b/eth/backend.go
@@ -637,7 +637,8 @@ func (s *Ethereum) Engine() consensus.Engine           { return s.engine }
 func (s *Ethereum) ChainDb() ethdb.Database            { return s.chainDb }
 func (s *Ethereum) IsListening() bool                  { return true } // Always listening
 func (s *Ethereum) Downloader() *downloader.Downloader { return s.handler.downloader }
-func (s *Ethereum) Synced() bool                       { return atomic.LoadUint32(&s.handler.acceptTxs) == 1 }
+func (s *Ethereum) Synced() bool                       { return atomic.LoadUint32(&s.handler.synced) == 1 }
+func (s *Ethereum) SetSynced()                         { s.handler.enableSyncedFeatures() }
 func (s *Ethereum) ArchiveMode() bool                  { return s.config.NoPruning }
 func (s *Ethereum) BloomIndexer() *core.ChainIndexer   { return s.bloomIndexer }
 
diff --git a/eth/downloader/downloader.go b/eth/downloader/downloader.go
index 345c6d98f..e7a9502e0 100644
--- a/eth/downloader/downloader.go
+++ b/eth/downloader/downloader.go
@@ -388,7 +388,9 @@ func (d *Downloader) synchronise(id string, hash common.Hash, td *big.Int, mode
 			// subsequent state reads, explicitly disable the trie database and state
 			// syncer is responsible to address and correct any state missing.
 			if d.blockchain.TrieDB().Scheme() == rawdb.PathScheme {
-				d.blockchain.TrieDB().Reset(types.EmptyRootHash)
+				if err := d.blockchain.TrieDB().Disable(); err != nil {
+					return err
+				}
 			}
 			// Snap sync uses the snapshot namespace to store potentially flakey data until
 			// sync completely heals and finishes. Pause snapshot maintenance in the mean
diff --git a/eth/handler.go b/eth/handler.go
index c6353e650..0c0e5f276 100644
--- a/eth/handler.go
+++ b/eth/handler.go
@@ -109,9 +109,9 @@ type handler struct {
 	networkID  uint64
 	forkFilter forkid.Filter // Fork ID filter, constant across the lifetime of the node
 
-	fastSync  uint32 // Flag whether fast sync is enabled (gets disabled if we already have blocks)
-	snapSync  uint32 // Flag whether fast sync should operate on top of the snap protocol
-	acceptTxs uint32 // Flag whether we're considered synchronised (enables transaction processing)
+	fastSync uint32 // Flag whether fast sync is enabled (gets disabled if we already have blocks)
+	snapSync uint32 // Flag whether fast sync should operate on top of the snap protocol
+	synced   uint32 // Flag whether we're considered synchronised (enables transaction processing)
 
 	checkpointNumber uint64      // Block number for the sync progress validator to cross reference
 	checkpointHash   common.Hash // Block hash for the sync progress validator to cross reference
@@ -183,17 +183,23 @@ func newHandler(config *handlerConfig) (*handler, error) {
 		fullBlock, fastBlock := h.chain.CurrentBlock(), h.chain.CurrentFastBlock()
 		if fullBlock.NumberU64() == 0 && fastBlock.NumberU64() > 0 {
 			h.fastSync = uint32(1)
-			log.Warn("Switch sync mode from full sync to fast sync")
+			log.Warn("Switch sync mode from full sync to fast sync", "reason", "snap sync incomplete")
+		} else if !h.chain.HasState(fullBlock.Root()) {
+			h.fastSync = uint32(1)
+			log.Warn("Switch sync mode from full sync to snap sync", "reason", "head state missing")
 		}
 	} else {
-		if h.chain.CurrentBlock().NumberU64() > 0 {
+		head := h.chain.CurrentBlock()
+		if head.NumberU64() > 0 {
 			// Print warning log if database is not empty to run fast sync.
 			log.Warn("Switch sync mode from fast sync to full sync")
 		} else {
 			// If fast sync was requested and our database is empty, grant it
 			h.fastSync = uint32(1)
+			log.Info("Enabled fast sync", "head", head.Number, "hash", head.Hash())
 			if config.Sync == downloader.SnapSync {
 				h.snapSync = uint32(1)
+				log.Info("Enabled snap sync", "head", head.Number, "hash", head.Hash())
 			}
 		}
 	}
@@ -237,15 +243,11 @@ func newHandler(config *handlerConfig) (*handler, error) {
 		// accept each others' blocks until a restart. Unfortunately we haven't figured
 		// out a way yet where nodes can decide unilaterally whether the network is new
 		// or not. This should be fixed if we figure out a solution.
-		if atomic.LoadUint32(&h.fastSync) == 1 {
-			log.Warn("Fast syncing, discarded propagated block", "number", blocks[0].Number(), "hash", blocks[0].Hash())
+		if atomic.LoadUint32(&h.synced) == 0 {
+			log.Warn("Syncing, discarded propagated block", "number", blocks[0].Number(), "hash", blocks[0].Hash())
 			return 0, nil
 		}
-		n, err := h.chain.InsertChain(blocks, sidecars)
-		if err == nil {
-			h.enableSyncedFeatures() // Mark initial sync done on any fetcher import
-		}
-		return n, err
+		return h.chain.InsertChain(blocks, sidecars)
 	}
 	h.blockFetcher = fetcher.NewBlockFetcher(false, nil, h.chain.GetBlockByHash, validator, h.chain.Engine().VerifyBlobHeader,
 		h.BroadcastBlock, heighter, nil, inserter, h.removePeer)
@@ -706,7 +708,18 @@ func (h *handler) voteBroadcastLoop() {
 // enableSyncedFeatures enables the post-sync functionalities when the initial
 // sync is finished.
 func (h *handler) enableSyncedFeatures() {
-	atomic.StoreUint32(&h.acceptTxs, 1)
+	atomic.StoreUint32(&h.synced, 1)
+
+	// If we were running fast/snap sync and it finished, disable doing another
+	// round on next sync cycle
+	if atomic.LoadUint32(&h.fastSync) == 1 {
+		log.Info("Fast sync complete, auto disabling")
+		atomic.StoreUint32(&h.fastSync, 0)
+	}
+	if atomic.LoadUint32(&h.snapSync) == 1 {
+		log.Info("Snap sync complete, auto disabling")
+		atomic.StoreUint32(&h.snapSync, 0)
+	}
 	if h.chain.TrieDB().Scheme() == rawdb.PathScheme {
 		h.chain.TrieDB().SetBufferSize(pathdb.DefaultBufferSize)
 	}
diff --git a/eth/handler_eth.go b/eth/handler_eth.go
index cfad1fdfa..6be4a5cd7 100644
--- a/eth/handler_eth.go
+++ b/eth/handler_eth.go
@@ -56,7 +56,7 @@ func (h *ethHandler) PeerInfo(id enode.ID) interface{} {
 // AcceptTxs retrieves whether transaction processing is enabled on the node
 // or if inbound transactions should simply be dropped.
 func (h *ethHandler) AcceptTxs() bool {
-	return atomic.LoadUint32(&h.acceptTxs) == 1
+	return atomic.LoadUint32(&h.synced) == 1
 }
 
 // Handle is invoked from a peer's message handler when it receives a new remote
diff --git a/eth/handler_eth_test.go b/eth/handler_eth_test.go
index 73f16a477..0661825ae 100644
--- a/eth/handler_eth_test.go
+++ b/eth/handler_eth_test.go
@@ -259,7 +259,7 @@ func testRecvTransactions(t *testing.T, protocol uint) {
 	handler := newTestHandler()
 	defer handler.close()
 
-	handler.handler.acceptTxs = 1 // mark synced to accept transactions
+	handler.handler.synced = 1 // mark synced to accept transactions
 
 	txs := make(chan core.NewTxsEvent)
 	sub := handler.txpool.SubscribeTransactions(txs, false)
@@ -408,7 +408,7 @@ func testTransactionPropagation(t *testing.T, protocol uint) {
 		sinks[i] = newTestHandler()
 		defer sinks[i].close()
 
-		sinks[i].handler.acceptTxs = 1 // mark synced to accept transactions
+		sinks[i].handler.synced = 1 // mark synced to accept transactions
 	}
 	// Interconnect all the sink handlers with the source handler
 	for i, sink := range sinks {
diff --git a/eth/sync.go b/eth/sync.go
index ecbbcc3e0..386132bf4 100644
--- a/eth/sync.go
+++ b/eth/sync.go
@@ -190,15 +190,24 @@ func (cs *chainSyncer) modeAndLocalHead() (downloader.SyncMode, *big.Int) {
 	}
 	// We are probably in full sync, but we might have rewound to before the
 	// fast sync pivot, check if we should reenable
+	head := cs.handler.chain.CurrentBlock()
 	if pivot := rawdb.ReadLastPivotNumber(cs.handler.database); pivot != nil {
-		if head := cs.handler.chain.CurrentBlock(); head.NumberU64() < *pivot {
+		if head.NumberU64() < *pivot {
 			block := cs.handler.chain.CurrentFastBlock()
 			td := cs.handler.chain.GetTd(block.Hash(), block.NumberU64())
 			return downloader.FastSync, td
 		}
 	}
+	// We are in a full sync, but the associated head state is missing. To complete
+	// the head state, forcefully rerun the snap sync. Note it doesn't mean the
+	// persistent state is corrupted, just mismatch with the head block.
+	if !cs.handler.chain.HasState(head.Root()) {
+		block := cs.handler.chain.CurrentFastBlock()
+		td := cs.handler.chain.GetTd(block.Hash(), block.NumberU64())
+		log.Info("Reenabled snap sync as chain is stateless")
+		return downloader.SnapSync, td
+	}
 	// Nope, we're really full syncing
-	head := cs.handler.chain.CurrentBlock()
 	td := cs.handler.chain.GetTd(head.Hash(), head.NumberU64())
 	return downloader.FullSync, td
 }
@@ -234,22 +243,14 @@ func (h *handler) doSync(op *chainSyncOp) error {
 	if err != nil {
 		return err
 	}
-	if atomic.LoadUint32(&h.fastSync) == 1 {
-		log.Info("Fast sync complete, auto disabling")
-		atomic.StoreUint32(&h.fastSync, 0)
-	}
-	if atomic.LoadUint32(&h.snapSync) == 1 {
-		log.Info("Snap sync complete, auto disabling")
-		atomic.StoreUint32(&h.snapSync, 0)
-	}
-	// If we've successfully finished a sync cycle and passed any required checkpoint,
-	// enable accepting transactions from the network.
+	h.enableSyncedFeatures()
+
 	head := h.chain.CurrentBlock()
 	if head.NumberU64() >= h.checkpointNumber {
 		// Checkpoint passed, sanity check the timestamp to have a fallback mechanism
 		// for non-checkpointed (number = 0) private networks.
 		if head.Time() >= uint64(time.Now().AddDate(0, -1, 0).Unix()) {
-			atomic.StoreUint32(&h.acceptTxs, 1)
+			atomic.StoreUint32(&h.synced, 1)
 		}
 	}
 	if head.NumberU64() > 0 {
diff --git a/miner/miner_test.go b/miner/miner_test.go
index fa652b27e..dd2fec27c 100644
--- a/miner/miner_test.go
+++ b/miner/miner_test.go
@@ -57,6 +57,7 @@ func (m *mockBackend) TxPool() *txpool.TxPool {
 }
 
 type testBlockChain struct {
+	root          common.Hash
 	statedb       *state.StateDB
 	gasLimit      uint64
 	chainHeadFeed *event.Feed
@@ -76,6 +77,10 @@ func (bc *testBlockChain) StateAt(common.Hash) (*state.StateDB, error) {
 	return bc.statedb, nil
 }
 
+func (bc *testBlockChain) HasState(root common.Hash) bool {
+	return bc.root == root
+}
+
 func (bc *testBlockChain) SubscribeChainHeadEvent(ch chan<- core.ChainHeadEvent) event.Subscription {
 	return bc.chainHeadFeed.Subscribe(ch)
 }
@@ -252,7 +257,7 @@ func createMiner(t *testing.T) (*Miner, *event.TypeMux) {
 		t.Fatalf("can't create new chain %v", err)
 	}
 	statedb, _ := state.New(bc.Genesis().Root(), bc.StateCache(), nil)
-	blockchain := &testBlockChain{statedb, 10000000, new(event.Feed)}
+	blockchain := &testBlockChain{bc.Genesis().Root(), statedb, 10000000, new(event.Feed)}
 
 	legacyPool := legacypool.New(testTxPoolConfig, bc.Config(), blockchain)
 	txPool, err := txpool.New(testTxPoolConfig.PriceLimit, blockchain, []txpool.SubPool{legacyPool})
diff --git a/trie/database.go b/trie/database.go
index 79ce83588..42eedc200 100644
--- a/trie/database.go
+++ b/trie/database.go
@@ -280,15 +280,27 @@ func (db *Database) Recoverable(root common.Hash) (bool, error) {
 	return pdb.Recoverable(root), nil
 }
 
-// Reset wipes all available journal from the persistent database and discard
-// all caches and diff layers. Using the given root to create a new disk layer.
+// Disable deactivates the database and invalidates all available state layers
+// as stale to prevent access to the persistent state, which is in the syncing
+// stage.
+//
 // It's only supported by path-based database and will return an error for others.
-func (db *Database) Reset(root common.Hash) error {
+func (db *Database) Disable() error {
+	pdb, ok := db.backend.(*pathdb.Database)
+	if !ok {
+		return errors.New("not supported")
+	}
+	return pdb.Disable()
+}
+
+// Enable activates database and resets the state tree with the provided persistent
+// state root once the state sync is finished.
+func (db *Database) Enable(root common.Hash) error {
 	pdb, ok := db.backend.(*pathdb.Database)
 	if !ok {
 		return errors.New("not supported")
 	}
-	return pdb.Reset(root)
+	return pdb.Enable(root)
 }
 
 // Journal commits an entire diff hierarchy to disk into a single journal entry.
diff --git a/trie/sync.go b/trie/sync.go
index a6338b8a8..2b257fe2f 100644
--- a/trie/sync.go
+++ b/trie/sync.go
@@ -25,6 +25,7 @@ import (
 	"github.com/ethereum/go-ethereum/core/rawdb"
 	"github.com/ethereum/go-ethereum/ethdb"
 	"github.com/ethereum/go-ethereum/log"
+	"github.com/ethereum/go-ethereum/metrics"
 )
 
 // ErrNotRequested is returned by the trie sync when it's requested to process a
@@ -40,6 +41,16 @@ var ErrAlreadyProcessed = errors.New("already processed")
 // memory if the node was configured with a significant number of peers.
 const maxFetchesPerDepth = 16384
 
+var (
+	// deletionGauge is the metric to track how many trie node deletions
+	// are performed in total during the sync process.
+	deletionGauge = metrics.NewRegisteredGauge("trie/sync/delete", nil)
+
+	// lookupGauge is the metric to track how many trie node lookups are
+	// performed to determine if node needs to be deleted.
+	lookupGauge = metrics.NewRegisteredGauge("trie/sync/lookup", nil)
+)
+
 // SyncPath is a path tuple identifying a particular trie node either in a single
 // trie (account) or a layered trie (account -> storage).
 //
@@ -75,9 +86,10 @@ func NewSyncPath(path []byte) SyncPath {
 
 // nodeRequest represents a scheduled or already in-flight trie node retrieval request.
 type nodeRequest struct {
-	hash common.Hash // Hash of the trie node to retrieve
-	path []byte      // Merkle path leading to this node for prioritization
-	data []byte      // Data content of the node, cached until all subtrees complete
+	hash    common.Hash // Hash of the trie node to retrieve
+	path    []byte      // Merkle path leading to this node for prioritization
+	data    []byte      // Data content of the node, cached until all subtrees complete
+	deletes [][]byte    // List of internal path segments for trie nodes to delete
 
 	parent   *nodeRequest // Parent state node referencing this entry
 	deps     int          // Number of dependencies before allowed to commit this node
@@ -107,17 +119,19 @@ type CodeSyncResult struct {
 // syncMemBatch is an in-memory buffer of successfully downloaded but not yet
 // persisted data items.
 type syncMemBatch struct {
-	nodes  map[string][]byte      // In-memory membatch of recently completed nodes
-	hashes map[string]common.Hash // Hashes of recently completed nodes
-	codes  map[common.Hash][]byte // In-memory membatch of recently completed codes
+	nodes   map[string][]byte      // In-memory membatch of recently completed nodes
+	hashes  map[string]common.Hash // Hashes of recently completed nodes
+	deletes map[string]struct{}    // List of paths for trie node to delete
+	codes   map[common.Hash][]byte // In-memory membatch of recently completed codes
 }
 
 // newSyncMemBatch allocates a new memory-buffer for not-yet persisted trie nodes.
 func newSyncMemBatch() *syncMemBatch {
 	return &syncMemBatch{
-		nodes:  make(map[string][]byte),
-		hashes: make(map[string]common.Hash),
-		codes:  make(map[common.Hash][]byte),
+		nodes:   make(map[string][]byte),
+		hashes:  make(map[string]common.Hash),
+		deletes: make(map[string]struct{}),
+		codes:   make(map[common.Hash][]byte),
 	}
 }
 
@@ -358,7 +372,7 @@ func (s *Sync) ProcessNode(result NodeSyncResult) error {
 // Commit flushes the data stored in the internal membatch out to persistent
 // storage, returning any occurred error.
 func (s *Sync) Commit(dbw ethdb.Batch) error {
-	// Dump the membatch into a database dbw
+	// Flush the pending node writes into database batch.
 	for path, value := range s.membatch.nodes {
 		owner, inner := ResolvePath([]byte(path))
 		rawdb.WriteTrieNode(dbw, owner, inner, s.membatch.hashes[path], value, s.scheme)
@@ -367,14 +381,21 @@ func (s *Sync) Commit(dbw ethdb.Batch) error {
 			s.bloom.Add(hash[:])
 		}
 	}
+	// Flush the pending node deletes into the database batch.
+	// Please note that each written and deleted node has a
+	// unique path, ensuring no duplication occurs.
+	for path := range s.membatch.deletes {
+		owner, inner := ResolvePath([]byte(path))
+		rawdb.DeleteTrieNode(dbw, owner, inner, common.Hash{} /* unused */, s.scheme)
+	}
+	// Flush the pending code writes into database batch.
 	for hash, value := range s.membatch.codes {
 		rawdb.WriteCode(dbw, hash, value)
 		if s.bloom != nil {
 			s.bloom.Add(hash[:])
 		}
 	}
-	// Drop the membatch data and return
-	s.membatch = newSyncMemBatch()
+	s.membatch = newSyncMemBatch() // reset the batch
 	return nil
 }
 
@@ -438,6 +459,39 @@ func (s *Sync) children(req *nodeRequest, object node) ([]*nodeRequest, error) {
 			node: node.Val,
 			path: append(append([]byte(nil), req.path...), key...),
 		}}
+		// Mark all internal nodes between shortNode and its **in disk**
+		// child as invalid. This is essential in the case of path mode
+		// scheme; otherwise, state healing might overwrite existing child
+		// nodes silently while leaving a dangling parent node within the
+		// range of this internal path on disk. This would break the
+		// guarantee for state healing.
+		//
+		// While it's possible for this shortNode to overwrite a previously
+		// existing full node, the other branches of the fullNode can be
+		// retained as they remain untouched and complete.
+		//
+		// This step is only necessary for path mode, as there is no deletion
+		// in hash mode at all.
+		if _, ok := node.Val.(hashNode); ok && s.scheme == rawdb.PathScheme {
+			owner, inner := ResolvePath(req.path)
+			for i := 1; i < len(key); i++ {
+				// While checking for a non-existent item in Pebble can be less efficient
+				// without a bloom filter, the relatively low frequency of lookups makes
+				// the performance impact negligible.
+				var exists bool
+				if owner == (common.Hash{}) {
+					exists = rawdb.ExistsAccountTrieNode(s.database, append(inner, key[:i]...))
+				} else {
+					exists = rawdb.ExistsStorageTrieNode(s.database, owner, append(inner, key[:i]...))
+				}
+				if exists {
+					req.deletes = append(req.deletes, key[:i])
+					deletionGauge.Inc(1)
+					log.Debug("Detected dangling node", "owner", owner, "path", append(inner, key[:i]...))
+				}
+			}
+			lookupGauge.Inc(int64(len(key) - 1))
+		}
 	case *fullNode:
 		for i := 0; i < 17; i++ {
 			if node.Children[i] != nil {
@@ -507,6 +561,11 @@ func (s *Sync) commitNodeRequest(req *nodeRequest) error {
 	s.membatch.nodes[string(req.path)] = req.data
 	s.membatch.hashes[string(req.path)] = req.hash
 
+	// Delete the internal nodes which are marked as invalid
+	for _, segment := range req.deletes {
+		path := append(req.path, segment...)
+		s.membatch.deletes[string(path)] = struct{}{}
+	}
 	delete(s.nodeReqs, string(req.path))
 	s.fetches[len(req.path)]--
 
diff --git a/trie/sync_test.go b/trie/sync_test.go
index b3fb10db1..a07dbccd8 100644
--- a/trie/sync_test.go
+++ b/trie/sync_test.go
@@ -76,31 +76,53 @@ func makeTestTrie(scheme string) (ethdb.Database, *Database, *SecureTrie, map[st
 
 // checkTrieContents cross references a reconstructed trie with an expected data
 // content map.
-func checkTrieContents(t *testing.T, db ethdb.Database, scheme string, root []byte, content map[string][]byte) {
+func checkTrieContents(t *testing.T, db ethdb.Database, scheme string, root []byte, content map[string][]byte, rawTrie bool) {
 	// Check root availability and trie contents
 	ndb := newTestDatabase(db, scheme)
-	trie, err := NewSecure(TrieID(common.BytesToHash(root)), ndb)
-	if err != nil {
-		t.Fatalf("failed to create trie at %x: %v", root, err)
-	}
-	if err := checkTrieConsistency(db, scheme, common.BytesToHash(root)); err != nil {
+	if err := checkTrieConsistency(db, scheme, common.BytesToHash(root), rawTrie); err != nil {
 		t.Fatalf("inconsistent trie at %x: %v", root, err)
 	}
+	type reader interface {
+		Get(key []byte) []byte
+	}
+	var r reader
+	if rawTrie {
+		trie, err := New(TrieID(common.BytesToHash(root)), ndb)
+		if err != nil {
+			t.Fatalf("failed to create trie at %x: %v", root, err)
+		}
+		r = trie
+	} else {
+		trie, err := NewSecure(TrieID(common.BytesToHash(root)), ndb)
+		if err != nil {
+			t.Fatalf("failed to create trie at %x: %v", root, err)
+		}
+		r = trie
+	}
 	for key, val := range content {
-		if have := trie.Get([]byte(key)); !bytes.Equal(have, val) {
+		if have := r.Get([]byte(key)); !bytes.Equal(have, val) {
 			t.Errorf("entry %x: content mismatch: have %x, want %x", key, have, val)
 		}
 	}
 }
 
 // checkTrieConsistency checks that all nodes in a trie are indeed present.
-func checkTrieConsistency(db ethdb.Database, scheme string, root common.Hash) error {
+func checkTrieConsistency(db ethdb.Database, scheme string, root common.Hash, rawTrie bool) error {
 	ndb := newTestDatabase(db, scheme)
-	trie, err := NewSecure(TrieID(root), ndb)
-	if err != nil {
-		return nil // Consider a non existent state consistent
+	var it NodeIterator
+	if rawTrie {
+		trie, err := New(TrieID(root), ndb)
+		if err != nil {
+			return nil // Consider a non existent state consistent
+		}
+		it = trie.MustNodeIterator(nil)
+	} else {
+		trie, err := NewSecure(TrieID(root), ndb)
+		if err != nil {
+			return nil // Consider a non existent state consistent
+		}
+		it = trie.MustNodeIterator(nil)
 	}
-	it := trie.MustNodeIterator(nil)
 	for it.Next(true) {
 	}
 	return it.Error()
@@ -211,7 +233,7 @@ func testIterativeSync(t *testing.T, count int, bypath bool, scheme string) {
 		}
 	}
 	// Cross check that the two tries are in sync
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData, false)
 }
 
 // Tests that the trie scheduler can correctly reconstruct the state even if only
@@ -276,7 +298,7 @@ func testIterativeDelayedSync(t *testing.T, scheme string) {
 		}
 	}
 	// Cross check that the two tries are in sync
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData, false)
 }
 
 // Tests that given a root hash, a trie can sync iteratively on a single thread,
@@ -346,7 +368,7 @@ func testIterativeRandomSync(t *testing.T, count int, scheme string) {
 		}
 	}
 	// Cross check that the two tries are in sync
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData, false)
 }
 
 // Tests that the trie scheduler can correctly reconstruct the state even if only
@@ -418,7 +440,7 @@ func testIterativeRandomDelayedSync(t *testing.T, scheme string) {
 		}
 	}
 	// Cross check that the two tries are in sync
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData, false)
 }
 
 // Tests that a trie sync will not request nodes multiple times, even if they
@@ -491,7 +513,7 @@ func testDuplicateAvoidanceSync(t *testing.T, scheme string) {
 		}
 	}
 	// Cross check that the two tries are in sync
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData, false)
 }
 
 // Tests that at any point in time during a sync, only complete sub-tries are in
@@ -577,7 +599,7 @@ func testIncompleteSync(t *testing.T, scheme string) {
 		nodeHash := addedHashes[i]
 		value := rawdb.ReadTrieNode(diskdb, owner, inner, nodeHash, scheme)
 		rawdb.DeleteTrieNode(diskdb, owner, inner, nodeHash, scheme)
-		if err := checkTrieConsistency(diskdb, srcDb.Scheme(), root); err == nil {
+		if err := checkTrieConsistency(diskdb, srcDb.Scheme(), root, false); err == nil {
 			t.Fatalf("trie inconsistency not caught, missing: %x", path)
 		}
 		rawdb.WriteTrieNode(diskdb, owner, inner, nodeHash, value, scheme)
@@ -653,7 +675,7 @@ func testSyncOrdering(t *testing.T, scheme string) {
 		}
 	}
 	// Cross check that the two tries are in sync
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData, false)
 
 	// Check that the trie nodes have been requested path-ordered
 	for i := 0; i < len(reqs)-1; i++ {
@@ -674,7 +696,7 @@ func syncWith(t *testing.T, root common.Hash, db ethdb.Database, srcDb *Database
 
 	// The code requests are ignored here since there is no code
 	// at the testing trie.
-	paths, nodes, _ := sched.Missing(1)
+	paths, nodes, _ := sched.Missing(0)
 	var elements []trieElement
 	for i := 0; i < len(paths); i++ {
 		elements = append(elements, trieElement{
@@ -708,7 +730,7 @@ func syncWith(t *testing.T, root common.Hash, db ethdb.Database, srcDb *Database
 		}
 		batch.Write()
 
-		paths, nodes, _ = sched.Missing(1)
+		paths, nodes, _ = sched.Missing(0)
 		elements = elements[:0]
 		for i := 0; i < len(paths); i++ {
 			elements = append(elements, trieElement{
@@ -734,7 +756,7 @@ func testSyncMovingTarget(t *testing.T, scheme string) {
 	// Create a destination trie and sync with the scheduler
 	diskdb := rawdb.NewMemoryDatabase()
 	syncWith(t, srcTrie.Hash(), diskdb, srcDb)
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), srcData, false)
 
 	// Push more modifications into the src trie, to see if dest trie can still
 	// sync with it(overwrite stale states)
@@ -758,7 +780,7 @@ func testSyncMovingTarget(t *testing.T, scheme string) {
 	srcTrie, _ = NewSecure(TrieID(root), srcDb)
 
 	syncWith(t, srcTrie.Hash(), diskdb, srcDb)
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), diff)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), diff, false)
 
 	// Revert added modifications from the src trie, to see if dest trie can still
 	// sync with it(overwrite reverted states)
@@ -782,5 +804,98 @@ func testSyncMovingTarget(t *testing.T, scheme string) {
 	srcTrie, _ = NewSecure(TrieID(root), srcDb)
 
 	syncWith(t, srcTrie.Hash(), diskdb, srcDb)
-	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), reverted)
+	checkTrieContents(t, diskdb, srcDb.Scheme(), srcTrie.Hash().Bytes(), reverted, false)
+}
+
+// Tests if state syncer can correctly catch up the pivot move.
+func TestPivotMove(t *testing.T) {
+	testPivotMove(t, rawdb.HashScheme, true)
+	testPivotMove(t, rawdb.HashScheme, false)
+	testPivotMove(t, rawdb.PathScheme, true)
+	testPivotMove(t, rawdb.PathScheme, false)
+}
+
+func testPivotMove(t *testing.T, scheme string, tiny bool) {
+	var (
+		srcDisk    = rawdb.NewMemoryDatabase()
+		srcTrieDB  = newTestDatabase(srcDisk, scheme)
+		srcTrie, _ = New(TrieID(types.EmptyRootHash), srcTrieDB)
+
+		deleteFn = func(key []byte, tr *Trie, states map[string][]byte) {
+			tr.Delete(key)
+			delete(states, string(key))
+		}
+		writeFn = func(key []byte, val []byte, tr *Trie, states map[string][]byte) {
+			if val == nil {
+				if tiny {
+					val = randBytes(4)
+				} else {
+					val = randBytes(32)
+				}
+			}
+			tr.Update(key, val)
+			states[string(key)] = common.CopyBytes(val)
+		}
+		copyStates = func(states map[string][]byte) map[string][]byte {
+			cpy := make(map[string][]byte)
+			for k, v := range states {
+				cpy[k] = v
+			}
+			return cpy
+		}
+	)
+	stateA := make(map[string][]byte)
+	writeFn([]byte{0x01, 0x23}, nil, srcTrie, stateA)
+	writeFn([]byte{0x01, 0x24}, nil, srcTrie, stateA)
+	writeFn([]byte{0x12, 0x33}, nil, srcTrie, stateA)
+	writeFn([]byte{0x12, 0x34}, nil, srcTrie, stateA)
+	writeFn([]byte{0x02, 0x34}, nil, srcTrie, stateA)
+	writeFn([]byte{0x13, 0x44}, nil, srcTrie, stateA)
+
+	rootA, nodesA, _ := srcTrie.Commit(false)
+	if err := srcTrieDB.Update(rootA, types.EmptyRootHash, 0, trienode.NewWithNodeSet(nodesA), nil); err != nil {
+		panic(err)
+	}
+	if err := srcTrieDB.Commit(rootA, false); err != nil {
+		panic(err)
+	}
+	// Create a destination trie and sync with the scheduler
+	destDisk := rawdb.NewMemoryDatabase()
+	syncWith(t, rootA, destDisk, srcTrieDB)
+	checkTrieContents(t, destDisk, scheme, srcTrie.Hash().Bytes(), stateA, true)
+
+	// Delete element to collapse trie
+	stateB := copyStates(stateA)
+	srcTrie, _ = New(TrieID(rootA), srcTrieDB)
+	deleteFn([]byte{0x02, 0x34}, srcTrie, stateB)
+	deleteFn([]byte{0x13, 0x44}, srcTrie, stateB)
+	writeFn([]byte{0x01, 0x24}, nil, srcTrie, stateB)
+
+	rootB, nodesB, _ := srcTrie.Commit(false)
+	if err := srcTrieDB.Update(rootB, rootA, 0, trienode.NewWithNodeSet(nodesB), nil); err != nil {
+		panic(err)
+	}
+	if err := srcTrieDB.Commit(rootB, false); err != nil {
+		panic(err)
+	}
+	syncWith(t, rootB, destDisk, srcTrieDB)
+	checkTrieContents(t, destDisk, scheme, srcTrie.Hash().Bytes(), stateB, true)
+
+	// Add elements to expand trie
+	stateC := copyStates(stateB)
+	srcTrie, _ = New(TrieID(rootB), srcTrieDB)
+
+	writeFn([]byte{0x01, 0x24}, stateA[string([]byte{0x01, 0x24})], srcTrie, stateC)
+	writeFn([]byte{0x02, 0x34}, nil, srcTrie, stateC)
+	writeFn([]byte{0x13, 0x44}, nil, srcTrie, stateC)
+
+	rootC, nodesC, _ := srcTrie.Commit(false)
+	if err := srcTrieDB.Update(rootC, rootB, 0, trienode.NewWithNodeSet(nodesC), nil); err != nil {
+		panic(err)
+	}
+	if err := srcTrieDB.Commit(rootC, false); err != nil {
+		panic(err)
+	}
+	syncWith(t, rootC, destDisk, srcTrieDB)
+	checkTrieContents(t, destDisk, scheme, srcTrie.Hash().Bytes(), stateC, true)
 }
diff --git a/trie/triedb/pathdb/database.go b/trie/triedb/pathdb/database.go
index f02f2164e..873418c9f 100644
--- a/trie/triedb/pathdb/database.go
+++ b/trie/triedb/pathdb/database.go
@@ -127,7 +127,8 @@ type Database struct {
 	// readOnly is the flag whether the mutation is allowed to be applied.
 	// It will be set automatically when the database is journaled during
 	// the shutdown to reject all following unexpected mutations.
-	readOnly   bool                     // Indicator if database is opened in read only mode
+	readOnly   bool                     // Flag if database is opened in read only mode
+	waitSync   bool                     // Flag if database is deactivated due to initial state sync
 	bufferSize int                      // Memory allowance (in bytes) for caching dirty nodes
 	config     *Config                  // Configuration for database
 	diskdb     ethdb.Database           // Persistent storage for matured trie nodes
@@ -176,6 +177,12 @@ func New(diskdb ethdb.Database, config *Config) *Database {
 			log.Warn("Truncated extra state histories from freezer", "count", pruned)
 		}
 	}
+	// Disable database in case node is still in the initial state sync stage.
+	if rawdb.ReadSnapSyncStatusFlag(diskdb) == rawdb.StateSyncRunning && !db.readOnly {
+		if err := db.Disable(); err != nil {
+			log.Crit("Failed to disable database", "err", err) // impossible to happen
+		}
+	}
 	log.Warn("Path-based state scheme is an experimental feature")
 	return db
 }
@@ -201,9 +208,9 @@ func (db *Database) Update(root common.Hash, parentRoot common.Hash, block uint6
 	db.lock.Lock()
 	defer db.lock.Unlock()
 
-	// Short circuit if the database is in read only mode.
-	if db.readOnly {
-		return errSnapshotReadOnly
+	// Short circuit if the mutation is not allowed.
+	if err := db.modifyAllowed(); err != nil {
+		return err
 	}
 	if err := db.tree.add(root, parentRoot, block, nodes, states); err != nil {
 		return err
@@ -228,45 +235,59 @@ func (db *Database) Commit(root common.Hash, report bool) error {
 	db.lock.Lock()
 	defer db.lock.Unlock()
 
-	// Short circuit if the database is in read only mode.
-	if db.readOnly {
-		return errSnapshotReadOnly
+	// Short circuit if the mutation is not allowed.
+	if err := db.modifyAllowed(); err != nil {
+		return err
 	}
 	return db.tree.cap(root, 0)
 }
 
-// Reset rebuilds the database with the specified state as the base.
-//
-//   - if target state is empty, clear the stored state and all layers on top
-//   - if target state is non-empty, ensure the stored state matches with it
-//     and clear all other layers on top.
-func (db *Database) Reset(root common.Hash) error {
+// Disable deactivates the database and invalidates all available state layers
+// as stale to prevent access to the persistent state, which is in the syncing
+// stage.
+func (db *Database) Disable() error {
 	db.lock.Lock()
 	defer db.lock.Unlock()
 
 	// Short circuit if the database is in read only mode.
 	if db.readOnly {
-		return errSnapshotReadOnly
+		return errDatabaseReadOnly
 	}
-	batch := db.diskdb.NewBatch()
-	root = types.TrieRootHash(root)
-	if root == types.EmptyRootHash {
-		// Empty state is requested as the target, nuke out
-		// the root node and leave all others as dangling.
-		rawdb.DeleteAccountTrieNode(batch, nil)
-	} else {
-		// Ensure the requested state is existent before any
-		// action is applied.
-		_, hash := rawdb.ReadAccountTrieNode(db.diskdb, nil)
-		if hash != root {
-			return fmt.Errorf("state is mismatched, local: %x, target: %x", hash, root)
-		}
+	// Prevent duplicated disable operation.
+	if db.waitSync {
+		log.Error("Reject duplicated disable operation")
+		return nil
 	}
-	// Mark the disk layer as stale before applying any mutation.
+	db.waitSync = true
+
+	// Mark the disk layer as stale to prevent access to persistent state.
 	db.tree.bottom().markStale()
 
+	// Write the initial sync flag to persist it across restarts.
+	rawdb.WriteSnapSyncStatusFlag(db.diskdb, rawdb.StateSyncRunning)
+	log.Info("Disabled trie database due to state sync")
+	return nil
+}
+
+// Enable activates database and resets the state tree with the provided persistent
+// state root once the state sync is finished.
+func (db *Database) Enable(root common.Hash) error {
+	db.lock.Lock()
+	defer db.lock.Unlock()
+
+	// Short circuit if the database is in read only mode.
+	if db.readOnly {
+		return errDatabaseReadOnly
+	}
+	// Ensure the provided state root matches the stored one.
+	root = types.TrieRootHash(root)
+	_, stored := rawdb.ReadAccountTrieNode(db.diskdb, nil)
+	if stored != root {
+		return fmt.Errorf("state root mismatch: stored %x, synced %x", stored, root)
+	}
 	// Drop the stale state journal in persistent database and
 	// reset the persistent state id back to zero.
+	batch := db.diskdb.NewBatch()
 	rawdb.DeleteTrieJournal(batch)
 	rawdb.WritePersistentStateID(batch, 0)
 	if err := batch.Write(); err != nil {
@@ -285,8 +306,11 @@ func (db *Database) Reset(root common.Hash) error {
 
 	// Re-construct a new disk layer backed by persistent state
 	// with **empty clean cache and node buffer**.
-	dl := newDiskLayer(root, 0, db, nil, newNodeBuffer(db.bufferSize, nil, 0))
-	db.tree.reset(dl)
+	db.tree.reset(newDiskLayer(root, 0, db, nil, newNodeBuffer(db.bufferSize, nil, 0)))
+
+	// Re-enable the database as the final step.
+	db.waitSync = false
+	rawdb.WriteSnapSyncStatusFlag(db.diskdb, rawdb.StateSyncFinished)
 	log.Info("Rebuilt trie database", "root", root)
 	return nil
 }
@@ -299,8 +323,11 @@ func (db *Database) Recover(root common.Hash, loader triestate.TrieLoader) error
 	defer db.lock.Unlock()
 
 	// Short circuit if rollback operation is not supported.
-	if db.readOnly || db.freezer == nil {
-		return errors.New("state rollback is not supported")
+	if err := db.modifyAllowed(); err != nil {
+		return err
+	}
+	if db.freezer == nil {
+		return errors.New("state rollback is non-supported")
 	}
 
 	// Short circuit if the target state is not recoverable.
@@ -438,3 +465,15 @@ func (db *Database) Close() error {
 	}
 	return db.freezer.Close()
 }
+
+// modifyAllowed returns the indicator if mutation is allowed. This function
+// assumes the db.lock is already held.
+func (db *Database) modifyAllowed() error {
+	if db.readOnly {
+		return errDatabaseReadOnly
+	}
+	if db.waitSync {
+		return errDatabaseWaitSync
+	}
+	return nil
+}
diff --git a/trie/triedb/pathdb/database_test.go b/trie/triedb/pathdb/database_test.go
index 37d92ac26..1eaaef691 100644
--- a/trie/triedb/pathdb/database_test.go
+++ b/trie/triedb/pathdb/database_test.go
@@ -440,38 +440,39 @@ func TestDatabaseRecoverable(t *testing.T) {
 	}
 }
 
-func TestReset(t *testing.T) {
-	var (
-		tester = newTester(t)
-		index  = tester.bottomIndex()
-	)
+func TestDisable(t *testing.T) {
+	tester := newTester(t)
 	defer tester.release()
 
-	// Reset database to unknown target, should reject it
-	if err := tester.db.Reset(testutil.RandomHash()); err == nil {
-		t.Fatal("Failed to reject invalid reset")
+	_, stored := rawdb.ReadAccountTrieNode(tester.db.diskdb, nil)
+	if err := tester.db.Disable(); err != nil {
+		t.Fatal("Failed to deactivate database")
 	}
-	// Reset database to state persisted in the disk
-	if err := tester.db.Reset(types.EmptyRootHash); err != nil {
-		t.Fatalf("Failed to reset database %v", err)
+	if err := tester.db.Enable(types.EmptyRootHash); err == nil {
+		t.Fatalf("Invalid activation should be rejected")
 	}
+	if err := tester.db.Enable(stored); err != nil {
+		t.Fatal("Failed to activate database")
+	}
+
 	// Ensure journal is deleted from disk
 	if blob := rawdb.ReadTrieJournal(tester.db.diskdb); len(blob) != 0 {
 		t.Fatal("Failed to clean journal")
 	}
 	// Ensure all trie histories are removed
-	for i := 0; i <= index; i++ {
-		_, err := readHistory(tester.db.freezer, uint64(i+1))
-		if err == nil {
-			t.Fatalf("Failed to clean state history, index %d", i+1)
-		}
+	n, err := tester.db.freezer.Ancients()
+	if err != nil {
+		t.Fatal("Failed to clean state history")
+	}
+	if n != 0 {
+		t.Fatal("Failed to clean state history")
 	}
 	// Verify layer tree structure, single disk layer is expected
 	if tester.db.tree.len() != 1 {
 		t.Fatalf("Extra layer kept %d", tester.db.tree.len())
 	}
-	if tester.db.tree.bottom().rootHash() != types.EmptyRootHash {
-		t.Fatalf("Root hash is not matched exp %x got %x", types.EmptyRootHash, tester.db.tree.bottom().rootHash())
+	if tester.db.tree.bottom().rootHash() != stored {
+		t.Fatalf("Root hash is not matched exp %x got %x", stored, tester.db.tree.bottom().rootHash())
 	}
 }
 
diff --git a/trie/triedb/pathdb/errors.go b/trie/triedb/pathdb/errors.go
index f503a9c49..450cbaa4b 100644
--- a/trie/triedb/pathdb/errors.go
+++ b/trie/triedb/pathdb/errors.go
@@ -24,9 +24,13 @@ import (
 )
 
 var (
-	// errSnapshotReadOnly is returned if the database is opened in read only mode
-	// and mutation is requested.
-	errSnapshotReadOnly = errors.New("read only")
+	// errDatabaseReadOnly is returned if the database is opened in read only mode
+	// to prevent any mutation.
+	errDatabaseReadOnly = errors.New("read only")
+
+	// errDatabaseWaitSync is returned if the initial state sync is not completed
+	// yet and database is disabled to prevent accessing state.
+	errDatabaseWaitSync = errors.New("waiting for sync")
 
 	// errSnapshotStale is returned from data accessors if the underlying layer
 	// layer had been invalidated due to the chain progressing forward far enough
diff --git a/trie/triedb/pathdb/journal.go b/trie/triedb/pathdb/journal.go
index f6e1854de..ec7d1d680 100644
--- a/trie/triedb/pathdb/journal.go
+++ b/trie/triedb/pathdb/journal.go
@@ -368,7 +368,7 @@ func (db *Database) Journal(root common.Hash) error {
 
 	// Short circuit if the database is in read only mode.
 	if db.readOnly {
-		return errSnapshotReadOnly
+		return errDatabaseReadOnly
 	}
 	// Firstly write out the metadata of journal
 	journal := new(bytes.Buffer)