test(rivetkit): simplify sqlite efficiency coverage

[NathanFlurry]

No description provided.

[NathanFlurry]

Stack for rivet-dev/rivet

• #5452 refactor(rivetkit): expose schedule intervals as durations
• #5450 perf(rivetkit): index named queue reads
• #5449 refactor(rivetkit): use unitless schedule duration names
• #5451 perf(rivetkit): bound queue reads and history pruning
• #5448 refactor(rivetkit): collapse unpublished internal schema
• #5446 refactor(rivetkit): centralize internal sqlite storage
• #5445 test(rivetkit): simplify sqlite efficiency coverage 👈
• #5443 perf(rivetkit): verify and optimize internal sqlite queries
• #5427 fix(rivetkit): clean up cron history efficiently
• #5439 docs(rivetkit): remove redundant overdue schedule note
• #5440 docs(rivetkit): use scheduling code snippets
• #5442 fix(rivetkit): configure inspector protocol versions
• #5441 feat(rivetkit): add actor scheduling
• main

Get stack: forklift get 5445
Push local edits: forklift submit
Merge when ready: forklift merge 5445

_{change spmlntnp}

[claude]

Review: test(rivetkit): simplify sqlite efficiency coverage

Small, well-scoped PR: adds an explanatory comment to clear_table_bounded and trims scaffolding from sql_efficiency.rs. A couple of notes:

1. Removed regression guard has no replacement (medium)

The deleted sql_efficiency_catalog_has_stable_unique_ids_and_documented_scans test enforced two invariants that no longer exist anywhere in the suite:

• Uniqueness of query ids across the catalog (previously also checked against CORRECTNESS_ONLY_QUERIES/EXCLUDED_QUERIES, but the uniqueness check applied to query_catalog() itself too). A future duplicate id in query_catalog() will no longer be caught, it'll just silently produce a confusing/ambiguous failure message instead of a clear "duplicate id" assertion.
• Minimum inventory size (original_len >= 50), which acted as a tripwire against someone accidentally deleting query cases over time without adding coverage elsewhere.

Given rivetkit-rust/packages/rivetkit-core/CLAUDE.md's policy that this suite is the enforcement mechanism for "every new/modified internal SQL statement needs coverage," losing this tripwire is a real reduction in the safety net, not just dead weight. Worth confirming this loss was intentional, or restoring a lightweight uniqueness/count check.

2. New top-of-file comment doesn't cover everything the deleted EXCLUDED_QUERIES explained (low)

The new comment:

// This suite covers SQL owned by rivetkit-core. User-provided SQL, including
// Actor Runtime Socket requests, is intentionally outside its scope.

only accounts for the actor_runtime_socket.sql exclusion. The other three removed EXCLUDED_QUERIES entries (inspector.schema_catalog, inspector.table_count, inspector.table_rows) were excluded for different reasons (schema metadata enumeration; user-selected table capped at 500 rows), those are core-owned SQL, not "user-provided SQL," so the new comment doesn't actually justify why they're absent from query_catalog(). A future reader auditing coverage has no way to tell whether these are intentionally excluded or just missing.

The 13 former CORRECTNESS_ONLY_QUERIES (singleton upserts/point inserts) are on firmer ground since the general policy in CLAUDE.md ("simple inserts... need correctness... but no plan assertion") already covers that class, so dropping their per-query reasons is a more defensible simplification.

3. Minor: duplicated comment (nit)

internal_storage.rs: the new comment on clear_table_bounded (private helper) largely restates the existing comment on its only caller, clear_imported_storage, a few lines up:

// Delete bounded sets of rows in separate commits. A single `DELETE FROM`
// over a large interrupted import can itself exceed depot's dirty-page
// limit, permanently preventing the importer from recovering.
...
async fn clear_table_bounded(...) {
	// An interrupted legacy import can leave enough data that one `DELETE FROM`
	// exceeds depot's dirty-page commit limit. Delete estimated-size chunks in
	// separate transactions so cleanup can always make progress.

Consider keeping the rationale in one place (probably on clear_table_bounded, since that's where the chunking actually happens) to avoid drift if the reasoning changes later.

Other

• Verified no other references to the removed CORRECTNESS_ONLY_QUERIES/EXCLUDED_QUERIES/meta-test exist elsewhere in the tree, so the deletion doesn't leave dangling references.
• No security or performance concerns, this is test-only plus a comment change.

[railway-app]

🚅 Deployed to the rivet-pr-5445 environment in rivet-frontend

Service	Status	Web	Updated (UTC)
frontend-inspector	😴 Sleeping (View Logs)	Web	Jul 20, 2026 at 7:02 am
frontend-cloud	😴 Sleeping (View Logs)	Web	Jul 20, 2026 at 7:01 am
kitchen-sink	😴 Sleeping (View Logs)	Web	Jul 20, 2026 at 7:01 am
website	😴 Sleeping (View Logs)	Web	Jul 20, 2026 at 7:00 am
ladle	✅ Success (View Logs)	Web	Jul 20, 2026 at 6:53 am
mcp-hub	✅ Success (View Logs)	Web	Jul 20, 2026 at 6:51 am