diff --git a/op-node/flags/flags.go b/op-node/flags/flags.go
index dc02e027209..e61808f5942 100644
--- a/op-node/flags/flags.go
+++ b/op-node/flags/flags.go
@@ -179,6 +179,13 @@ var (
 		Value:    20,
 		Category: L1RPCCategory,
 	}
+	L1CacheSize = &cli.UintFlag{
+		Name: "l1.cache-size",
+		Usage: "Cache size for blocks, receipts and transactions. " +
+			"It's optional and a sane default of 3/2 the sequencing window size is used if this field is set to 0.",
+		EnvVars:  prefixEnvVars("L1_CACHE_SIZE"),
+		Category: L1RPCCategory,
+	}
 	L1HTTPPollInterval = &cli.DurationFlag{
 		Name:     "l1.http-poll-interval",
 		Usage:    "Polling interval for latest-block subscription when using an HTTP RPC provider. Ignored for other types of RPC endpoints.",
@@ -423,6 +430,7 @@ var optionalFlags = []cli.Flag{
 	L1RPCMaxBatchSize,
 	L1RPCMaxConcurrency,
 	L1HTTPPollInterval,
+	L1CacheSize,
 	VerifierL1Confs,
 	SequencerEnabledFlag,
 	SequencerStoppedFlag,
diff --git a/op-node/node/client.go b/op-node/node/client.go
index d8df28daebf..3796bb98d05 100644
--- a/op-node/node/client.go
+++ b/op-node/node/client.go
@@ -120,6 +120,12 @@ type L1EndpointConfig struct {
 	// It is recommended to use websockets or IPC for efficient following of the changing block.
 	// Setting this to 0 disables polling.
 	HttpPollInterval time.Duration
+
+	// CacheSize specifies the cache size for blocks, receipts and transactions. It's optional and a
+	// sane default of 3/2 the sequencing window size is used during Setup if this field is set to 0.
+	// Note that receipts and transactions are cached per block, which is why there's only one cache
+	// size to configure.
+	CacheSize uint
 }
 
 var _ L1EndpointSetup = (*L1EndpointConfig)(nil)
@@ -129,11 +135,14 @@ func (cfg *L1EndpointConfig) Check() error {
 		return fmt.Errorf("batch size is invalid or unreasonable: %d", cfg.BatchSize)
 	}
 	if cfg.RateLimit < 0 {
-		return fmt.Errorf("rate limit cannot be negative")
+		return fmt.Errorf("rate limit cannot be negative: %f", cfg.RateLimit)
 	}
 	if cfg.MaxConcurrency < 1 {
 		return fmt.Errorf("max concurrent requests cannot be less than 1, was %d", cfg.MaxConcurrency)
 	}
+	if cfg.CacheSize > 1_000_000 {
+		return fmt.Errorf("cache size is dangerously large: %d", cfg.CacheSize)
+	}
 	return nil
 }
 
@@ -146,14 +155,20 @@ func (cfg *L1EndpointConfig) Setup(ctx context.Context, log log.Logger, rollupCf
 		opts = append(opts, client.WithRateLimit(cfg.RateLimit, cfg.BatchSize))
 	}
 
-	l1Node, err := client.NewRPC(ctx, log, cfg.L1NodeAddr, opts...)
+	l1RPC, err := client.NewRPC(ctx, log, cfg.L1NodeAddr, opts...)
 	if err != nil {
 		return nil, nil, fmt.Errorf("failed to dial L1 address (%s): %w", cfg.L1NodeAddr, err)
 	}
-	rpcCfg := sources.L1ClientDefaultConfig(rollupCfg, cfg.L1TrustRPC, cfg.L1RPCKind)
-	rpcCfg.MaxRequestsPerBatch = cfg.BatchSize
-	rpcCfg.MaxConcurrentRequests = cfg.MaxConcurrency
-	return l1Node, rpcCfg, nil
+
+	var l1Cfg *sources.L1ClientConfig
+	if cfg.CacheSize > 0 {
+		l1Cfg = sources.L1ClientSimpleConfig(cfg.L1TrustRPC, cfg.L1RPCKind, int(cfg.CacheSize))
+	} else {
+		l1Cfg = sources.L1ClientDefaultConfig(rollupCfg, cfg.L1TrustRPC, cfg.L1RPCKind)
+	}
+	l1Cfg.MaxRequestsPerBatch = cfg.BatchSize
+	l1Cfg.MaxConcurrentRequests = cfg.MaxConcurrency
+	return l1RPC, l1Cfg, nil
 }
 
 // PreparedL1Endpoint enables testing with an in-process pre-setup RPC connection to L1
diff --git a/op-node/node/config.go b/op-node/node/config.go
index 5cbb5b9de16..9fb112eb39a 100644
--- a/op-node/node/config.go
+++ b/op-node/node/config.go
@@ -130,7 +130,7 @@ func (cfg *Config) LoadPersisted(log log.Logger) error {
 // Check verifies that the given configuration makes sense
 func (cfg *Config) Check() error {
 	if err := cfg.L1.Check(); err != nil {
-		return fmt.Errorf("l2 endpoint config error: %w", err)
+		return fmt.Errorf("l1 endpoint config error: %w", err)
 	}
 	if err := cfg.L2.Check(); err != nil {
 		return fmt.Errorf("l2 endpoint config error: %w", err)
diff --git a/op-node/node/node.go b/op-node/node/node.go
index f39dca958ae..bd5da091e3f 100644
--- a/op-node/node/node.go
+++ b/op-node/node/node.go
@@ -188,13 +188,13 @@ func (n *OpNode) initTracer(ctx context.Context, cfg *Config) error {
 }
 
 func (n *OpNode) initL1(ctx context.Context, cfg *Config) error {
-	l1Node, rpcCfg, err := cfg.L1.Setup(ctx, n.log, &cfg.Rollup)
+	l1RPC, l1Cfg, err := cfg.L1.Setup(ctx, n.log, &cfg.Rollup)
 	if err != nil {
 		return fmt.Errorf("failed to get L1 RPC client: %w", err)
 	}
 
 	n.l1Source, err = sources.NewL1Client(
-		client.NewInstrumentedRPC(l1Node, &n.metrics.RPCMetrics.RPCClientMetrics), n.log, n.metrics.L1SourceCache, rpcCfg)
+		client.NewInstrumentedRPC(l1RPC, &n.metrics.RPCMetrics.RPCClientMetrics), n.log, n.metrics.L1SourceCache, l1Cfg)
 	if err != nil {
 		return fmt.Errorf("failed to create L1 source: %w", err)
 	}
diff --git a/op-node/service.go b/op-node/service.go
index d3e9c29b3c6..7df2ab9a121 100644
--- a/op-node/service.go
+++ b/op-node/service.go
@@ -160,6 +160,7 @@ func NewL1EndpointConfig(ctx *cli.Context) *node.L1EndpointConfig {
 		BatchSize:        ctx.Int(flags.L1RPCMaxBatchSize.Name),
 		HttpPollInterval: ctx.Duration(flags.L1HTTPPollInterval.Name),
 		MaxConcurrency:   ctx.Int(flags.L1RPCMaxConcurrency.Name),
+		CacheSize:        ctx.Uint(flags.L1CacheSize.Name),
 	}
 }
 
diff --git a/op-service/sources/l1_client.go b/op-service/sources/l1_client.go
index 0f1ed7ebcf7..b69429d14a2 100644
--- a/op-service/sources/l1_client.go
+++ b/op-service/sources/l1_client.go
@@ -24,25 +24,20 @@ func L1ClientDefaultConfig(config *rollup.Config, trustRPC bool, kind RPCProvide
 }
 
 func L1ClientSimpleConfig(trustRPC bool, kind RPCProviderKind, cacheSize int) *L1ClientConfig {
-	span := cacheSize
-	if span > 1000 { // sanity cap. If a large sequencing window is configured, do not make the cache too large
-		span = 1000
-	}
 	return &L1ClientConfig{
 		EthClientConfig: EthClientConfig{
 			// receipts and transactions are cached per block
-			ReceiptsCacheSize:     span,
-			TransactionsCacheSize: span,
-			HeadersCacheSize:      span,
-			PayloadsCacheSize:     span,
+			ReceiptsCacheSize:     cacheSize,
+			TransactionsCacheSize: cacheSize,
+			HeadersCacheSize:      cacheSize,
+			PayloadsCacheSize:     cacheSize,
 			MaxRequestsPerBatch:   20, // TODO: tune batch param
 			MaxConcurrentRequests: 10,
 			TrustRPC:              trustRPC,
 			MustBePostMerge:       false,
 			RPCProviderKind:       kind,
 			MethodResetDuration:   time.Minute,
-			// Not bounded by span, to cover find-sync-start range fully for speedy recovery after errors.
-			BlockRefsCacheSize: cacheSize,
+			BlockRefsCacheSize:    cacheSize,
 		},
 	}
 }