• fix(epoxy): use vbare #4616
• fix: dont upsert for ensure if missing #4615
• fix(envoy): require runner config for ws conn #4612
• fix(pb): backport serverless fix #4611
• fix(pb): optimize actor key allocation #4610
• fix: move runner configs to epoxy #4605
• chore(epoxy): clean up kv get #4604
• fix(pb-envoy): reduce round trips for ws connection #4603
• fix(envoy-client): fix handling actor event ack #4599 👈 (View in Graphite)
• fix(envoy-client): add tls #4595 : 1 other dependent PR (#4597 )
• chore: fix driver test suite #4591
• chore: tunnel auth #4588
• chore: remove udb as a dependency of envoy-client #4587
• chore: revert actor v2 hack #4586
• fix: fix pool metadata protocol version refresh, rivetkit native ws #4585
• feat: US-001 - Define SqliteKv trait in rivetkit-sqlite-native #4584
• feat(rust/engine-runner): minimal rust engine runner #4289
• chore: delete kv channel #4583
• feat(engine): rust runner #4582
• chore: remove non-engine drivers #4574 : 1 other dependent PR (#4581 )
• feat: dynamic actors #4397
• fix: driver test suite & sqlite reliability #4559
• fix: remove serverless token #4572
• fix: misc token fixes #4569
• fix(pb): clean up actor stop decision handling #4566
• fix(envoy): use global instance, add signal handlers #4565
• chore: misc fixes, add pb snapshot test #4561
• fix(cache): make in memory cache global #4560
• main

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

Code Review: PR #4599 — fix(envoy-client): fix handling actor event ack

PR Status: MERGED
Author: MasterPtato (Max)
Net change: +35 / -1671 (mostly deletions of old test/load-test scripts)

Summary of Changes

1. engine/sdks/rust/envoy-client/src/events.rs — Bug fix: properly remove actor entries from the context map after a stop event is acknowledged, and simplify the ack handler.
2. engine/sdks/rust/envoy-client/src/commands.rs + envoy.rs — Add received_stop: bool field to ActorEntry to track whether Rivet sent a stop command.
3. rivetkit-typescript/packages/rivetkit-native/wrapper.js — Polyfill CloseEvent for Node.js < v22.
4. rivetkit-typescript/packages/rivetkit/src/serverless/router.ts — Fix: prefer config.token over the ephemeral start token in the runner config.
5. rivetkit-typescript/packages/sqlite-native/Cargo.toml — Remove [profile.release] lto = true that was placed in the wrong crate.
6. scripts/tests/actor_spam.ts, scripts/tests/load-test/ — Delete stale test scripts.

Core Bug Fix (events.rs) — Positive Assessment

The primary fix is logically correct. The old handle_ack_events had a two-level cleanup: trimming acknowledged events from history, then removing the entry if history was empty and the actor handle was closed. This created a bug where actors stopped by Rivet would linger in ctx.actors if the ack arrived before the ActorStateStopped event. The new approach is cleaner:

• Actor removal now happens immediately when an ActorStateStopped event is received and a stop command was previously sent (received_stop == true). This is deterministic and event-driven.
• handle_ack_events is simplified to only retain unacknowledged events, with no cleanup responsibility.

Issue 1 — Actor entry removal uses actor_id key without considering generation (Medium)

ctx.actors.remove(&event.checkpoint.actor_id) removes the entire outer map entry for actor_id (all generations), rather than removing only the specific generation (event.checkpoint.generation). If two generations of the same actor ever coexist in ctx.actors, stopping one would silently evict the other. The previous code was more precise. Suggested fix:

if let Some(gens) = ctx.actors.get_mut(&event.checkpoint.actor_id) {
    gens.remove(&event.checkpoint.generation);
    if gens.is_empty() {
        ctx.actors.remove(&event.checkpoint.actor_id);
    }
}

Issue 2 — Actor self-stop leaves entry in ctx.actors until lost-timeout fires (Low)

The new handle_ack_events never removes entries. Cleanup now relies solely on handle_send_events observing ActorStateStopped with received_stop == true. Actors that self-stopped without a prior CommandStopActor are never removed from ctx.actors until the lost_timeout path fires. This is exactly the memory leak scenario documented in the TODO comment, but it also applies to normal actor self-termination without a prior stop command.

Issue 3 — Deferred work item not tracked (Low)

The TODO comment describes a known memory leak scenario (envoy disconnect, actor self-stops, envoy reconnects, stop event resent but entry never cleaned up). Per CLAUDE.md conventions, deferred work items with context should be tracked in .agent/todo/*.md rather than left only as inline comments.

serverless/router.ts Fix — Positive

The change from token to config.token ?? token is correct. The previous code always overwrote config.token with the ephemeral start-request token, breaking setups where a static application token should take precedence. This aligns runnerConfig with the existing pattern in clientConfig.

wrapper.js CloseEvent Polyfill — Looks Good

Correct and minimal. Node.js added CloseEvent as a global in v22, so the guard on typeof CloseEvent === 'undefined' is appropriate. The implementation matches the browser CloseEvent spec for the properties used (code, reason, wasClean).

sqlite-native/Cargo.toml — LTO Removal

Correct. Per Cargo semantics, [profile.*] in a non-root crate is silently ignored — it only applies at the workspace root. This was dead configuration.

CLAUDE.md Convention Compliance

All new Rust code uses hard tabs, tracing structured logging (no println!/eprintln!), and follows existing patterns. Comment style is sentence-case and complete. No new dependencies introduced. Clean.

Summary

The PR is a net improvement. The original ack handler had a real bug where event history trimming could prevent cleanup and cause duplicate resends on reconnect. The new event-driven approach is cleaner. Issue 1 (generation-scoped removal) is the most important thing to address in a follow-up.

Merge activity

• Apr 13, 12:21 AM UTC: A user started a stack merge that includes this pull request via Graphite.
• Apr 13, 12:27 AM UTC: Graphite rebased this pull request as part of a merge.
• Apr 13, 12:27 AM UTC: @MasterPtato merged this pull request with Graphite.