diff --git a/src/v2/compiler/02_parse.dag b/src/v2/compiler/02_parse.dag
index cfffde79f34..0901d8b11f6 100644
--- a/src/v2/compiler/02_parse.dag
+++ b/src/v2/compiler/02_parse.dag
@@ -1762,18 +1762,18 @@ fn grammar_validate_for_parse(root: GrammarRoot, at: Locus) -> Outcome<Validated
   }
 }
 
-fn parse_sync_step(acc: ParseSyncState, tok: Token, sync_tokens: Set<Symbol>) -> ParseSyncState {
+fn parse_sync_step(acc: ParseSyncState, tok: Token, sync_tokens: List<Symbol>) -> ParseSyncState {
   match acc.found_sync {
     true => ParseSyncState { found_sync: true, result: list_snoc_item(xs: acc.result, item: tok) }
     false =>
-      match sync_tokens.member(tok.class) {
+      match contains(xs: sync_tokens, item: tok.class, eq: fn(a, b) { a == b }) {
         true => ParseSyncState { found_sync: true, result: list_snoc_item(xs: std.algebra.Empty, item: tok) }
         false => ParseSyncState { found_sync: false, result: std.algebra.Empty }
       }
   }
 }
 
-fn parse_skip_to_sync(tokens: List<Token>, sync_tokens: Set<Symbol>) -> ParseSyncState {
+fn parse_skip_to_sync(tokens: List<Token>, sync_tokens: List<Symbol>) -> ParseSyncState {
   fold_list(
     xs: tokens,
     empty: ParseSyncState { found_sync: false, result: std.algebra.Empty },
diff --git a/src/v2/extdeps/languages/dag.dag b/src/v2/extdeps/languages/dag.dag
index a161fafaafc..77a160cf5ba 100644
--- a/src/v2/extdeps/languages/dag.dag
+++ b/src/v2/extdeps/languages/dag.dag
@@ -3456,17 +3456,15 @@ fn dag_lex_token_symbol(sym: Symbol) -> Bool {
   )
 }
 
-fn dag_sync_tokens() -> Set<Symbol> {
-  Set {
-    member: fn(sym) {
-      sym == ^dag_token_kw_module ||
-      sym == ^dag_token_kw_import ||
-      sym == ^dag_token_kw_alias ||
-      sym == ^dag_token_kw_type ||
-      sym == ^dag_token_kw_data ||
-      sym == ^dag_token_kw_fn
-    }
-  }
+fn dag_sync_tokens() -> List<Symbol> {
+  [
+    ^dag_token_kw_module,
+    ^dag_token_kw_import,
+    ^dag_token_kw_alias,
+    ^dag_token_kw_type,
+    ^dag_token_kw_data,
+    ^dag_token_kw_fn
+  ]
 }
 
 
diff --git a/src/v2/std/grammar.dag b/src/v2/std/grammar.dag
index 47a9f1f5965..f8a1fd550f4 100644
--- a/src/v2/std/grammar.dag
+++ b/src/v2/std/grammar.dag
@@ -289,10 +289,27 @@ type GrammarProduction {
   emitted: Node
 }
 
+// SYNC TOKENS ARE AN ENUMERATION, NOT A PREDICATE.
+//
+// A grammar's error-recovery sync tokens are finite and known at authoring time — every producer in
+// the corpus writes either an empty set or a literal list of keyword classes. Carrying them as
+// `Set<Symbol>` made them a POINTWISE-POWER characteristic function (`std.types` grounds
+// `Set<element> = PointwisePower<element>`), which is the right grounding for a set in general and
+// the wrong carrier for six literal keywords: a characteristic function cannot be enumerated, so
+// the whole grammar value became unhashable through this one field.
+//
+// That was not a theoretical loss. `prepare_grammar` is memoized cross-claim on
+// (declaration identity, content hash of the grammar), and the interpreter's content-hash walk
+// BAILS on a closure — so the memo could never form its key, never store, and never hit, for every
+// language in the corpus and not only `.dag`. The measured consequence was a full ~12s grammar
+// preparation on every parse, including of an empty module.
+//
+// The field is therefore the enumeration it always was. `Set` keeps its grounding untouched; what
+// changed is that this field stopped borrowing a power-set to carry a list.
 type GrammarRoot {
   start: Symbol
   productions: List<GrammarProduction>
-  sync_tokens: Set<Symbol>
+  sync_tokens: List<Symbol>
 }
 
 type ValidatedGrammarRoot {
@@ -347,8 +364,8 @@ fn grammar_named_edge(name: Symbol, target: Node) -> Edge {
   }
 }
 
-fn grammar_empty_sync_tokens() -> Set<Symbol> {
-  Set { member: fn(_) { false } }
+fn grammar_empty_sync_tokens() -> List<Symbol> {
+  []
 }
 
 fn grammar_empty_node() -> Node {