• fix(test): stabilize lifecycle, sleep, queue, and run edge cases #4660 : 2 dependent PRs (#4664 , #4665 ) 👈 (View in Graphite)
• test(rivetkit): re-enable gateway URL and direct-registry coverage #4659
• fix(rivetkit): add lifecycle error retry and gateway HTTP routing #4666
• chore(engine): publish engine bases in ci #4649 : 2 other dependent PRs (#4650 , #4652 )
• chore: lockfile lefthook #4647
• fix(rivetkit): surface native sqlite kv errors to callers #4645
• fix(rivetkit): prevent sleep races during disconnect and db work #4644
• fix(rivetkit): isolate engine envoys and propagate startup failures #4643
• test(rivetkit): remove wasm sqlite-only fixtures and coverage #4642
• chore(sqlite-vfs): remove wasm sqlite packages and workspace wiring #4641
• chore(rivetkit): remove wasm sqlite runtime #4614
• fix(sqlite-native): restore native startup kv preload #4639
• fix(sqlite-native): delete metadata before chunk range #4638
• fix(sqlite-native): keep truncate cache coherent #4637
• fix(sqlite-vfs): use delete range for truncate cleanup #4636
• fix(sqlite-native): restore kv error hook #4635
• perf(sqlite-native): avoid cloning cached read chunks #4634
• perf(sqlite-native): gate kv operation labels #4633
• perf(sqlite-native): reuse read cache for partial writes #4632
• perf(sqlite-native): remove delete file existence probe #4631
• fix(sqlite-native): gate native read cache behind env var #4630
• fix(pegboard): isolate runner config dc lookup failures #4625 : 1 other dependent PR (#4626 )
• docs(pegboard): note runner config upsert split write risk #4624
• fix(pegboard): unpack global runner config keys correctly #4623
• fix(pegboard-gateway): enforce tunnel message state #4622
• fix(pegboard-runner): clear terminal tunnel routes #4621
• fix(pegboard): skip protocol version keys in runner pool backfill #4620
• fix(pegboard): create missing normal runner configs #4619
• fix(epoxy): reuse cached optimistic value #4618
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

PR Review: fix(test): stabilize lifecycle, sleep, queue, and run edge cases

Overview

This PR stabilizes a broad set of flaky driver tests by:

• Replacing fixed waitFor delays with vi.waitFor polling
• Relaxing exact count assertions where multiple cycles can occur
• Making lifecycle tests tolerate inherent create/destroy race conditions
• Introducing a lifecycle observer actor to verify events on destroyed actors
• Using c.queue.send() directly instead of self-referencing via client
• Fixing Response.json() compatibility across runtimes

The approach is sound overall. These changes reflect real non-determinism in the system rather than test implementation bugs.

Issues

isDynamicSandboxRuntime() is fragile and undocumented (inline-client.ts)

The process.cwd() === "/root" check is a platform-detection heuristic that will silently break if the sandbox working directory changes. The behavior divergence is significant: the sandbox path skips WebSocket entirely and uses direct action calls. This needs a comment explaining the constraint, and ideally should use an explicit environment variable or driver config flag instead.

destroyActor silently swallows all errors (actor-lifecycle.ts)

The final bare return at the end of the catch block swallows ALL unexpected errors from destroy(). The service_unavailable and destroy-race branches are intentional, but the catch-all fallthrough should throw err rather than silently returning, or have a comment explaining why every possible error is acceptable here.

Duplicate event listener in broadcast test (actor-sleep-db.ts)

Both connection.once("sleeping", ...) and connection.on("sleeping", ...) are registered for the same event. The on handler is redundant since the once handler already sets sleepingEventReceived = true, and the on listener leaks (never unsubscribed). Only once is needed.

instanceof errors.ActorAborted replaced with duck-typing without explanation (queue.ts, actor/instance/queue.ts)

Changing to error.group/code comparison is likely the right fix for cross-VM-context errors where instanceof fails, but there is no comment explaining why. This is non-obvious and will confuse future maintainers.

Minor Notes

• resolveActorId returning null: When resolve() throws a destroy race error, downstream assertions are silently skipped. The outcome is still valid but a comment would clarify intent.
• waitForConnectionOpen duck-typing: Manually typed rather than using the existing connection type. If the type is exported, prefer it.
• Removed log line in run.ts: Dropping c.log.info for tick count is fine but will make debugging tick test failures harder.
• lifecycle-hooks.ts API change: Moving connection params from .connect() to .getOrCreate(undefined, params) is a real API change; worth noting if it affects any public surface.
• Sequential mode in actor-run.ts: Adding .sequential is correct but a brief comment on why would help future readers.

Positive Changes

• The lifecycle observer pattern is a clean design for verifying lifecycle on actors that may be destroyed before assertions run.
• triggerSleepTwice directly validates idempotency of the sleep flag within a single actor turn.
• c.queue.send() replacing the self-client pattern is the correct fix.
• Timeout increases and vi.waitFor polling are the right stabilization approach for async tests.

Generated with Claude Code