• refactor(depot-client): split embedded depot transport #4970 : 2 dependent PRs (#4971 , #4972 )
• fix(rivetkit): decode legacy v4 actor schedule args #4976
• fix(engine): install rustls provider with pools #4975
• fix(rivetkit): defer initialized persist until state exists #4974
• test(rivetkit): cover failed preflight disconnect handling #4973 👈 (View in Graphite)
• fix(rivetkit): hide pending connections during preflight #4969
• WIP: more fuzz tests #4962
• fix(rivetkit): wrap onMigrate in savepoints #4960 : 1 other dependent PR (#4963 )
• fix(pegboard): align drain grace defaults with stop threshold #4958 : 1 other dependent PR (#4966 )
• fix(rivetkit): preserve bridge error messages #4959
• fix(rivetkit): fire abort signal on shutdown finalize #4957
• fix(rivetkit): use startup kv preload #4956
• fix(client): flatten skipReadyWait option #4955
• ci(publish): add manual preview publish flow #4954
• fix(depot): add ltx decode diagnostics #4953
• fix(universaldb): preserve plain set values #4952
• fix(universaldb): align versionstamp operand encoding #4951
• fix(rivetkit): export database provider types #4950
• wip: kitchen sink agentic loop finish #4949
• chore(rivetkit): note duplicate admission logic #4948
• test(rivetkit): cover bypass fetch header forwarding #4947
• feat(kitchen-sink): add actor probe controls #4946
• fix(gateway): split actor response wait errors #4945
• fix(guard): handle bypass preflight before query parse #4944
• feat(kitchen-sink): add mock agent event log copy #4943
• fix(kitchen-sink): default mock agent to local endpoint #4942
• test(rivetkit): enable driver lifecycle regressions #4941
• test(rivetkit): verify cached websocket db refs exceed grace #4940
• fix(rivetkit-core): keep sleeping actors reachable during grace #4939
• test(rivetkit): cover async websocket close db access #4938
• fix(rivetkit): keep actors awake until keepAwake work finishes #4937
• fix(rivetkit): persist native onSleep state on errors #4936
• fix(rivetkit-core): allow dispatch during sleep grace #4935
• fix(rivetkit): make duplicate sleep requests idempotent #4934
• main

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

Code Review: test(rivetkit): cover failed preflight disconnect handling

Overview

This PR adds test coverage for the scenario where a connection transport closes (or the preflight is rejected) during the preflight phase, verifying that onDisconnect is never called for connections that never fully established. Coverage spans two layers:

1. Rust unit test (rivetkit-core/tests/connection.rs): verifies that no ConnectionClosed event is emitted when a transport is closed mid-preflight.
2. TypeScript driver test (conn-preflight-disconnect.test.ts): verifies from the client side that onDisconnect callbacks aren't fired and rejected connections don't appear in visibleLabels.

Issues

Narrow test matrix (most important)

The new TS driver test is the only test in the suite that restricts runtimes:

{
    encodings: ["bare"],
    runtimes: ["wasm"],       // only test in the suite to restrict this
    sqliteBackends: ["remote"],
}

Every other driver test (e.g. conn-error-serialization.test.ts) passes no runtime restriction and runs across all configured runtimes. If the preflight disconnect behavior is expected to work correctly in both native and wasm runtimes — and based on the Rust-layer test it should be — the matrix should include native as well. If there's a known reason this only applies to wasm (e.g. the bug path only exists there), a comment would help future readers understand why.

Incomplete snapshot assertions

The test asserts that disconnectSnapshots is empty and visibleLabels contains only "primary", but doesn't verify:

• connectSnapshots has exactly one entry (confirming onConnect fired only for the primary connection, not for the rejected ones).
• beforeStarted / createStarted counts to confirm the preflight hooks did run (but rejected).

These would lock in the full contract of the behavior being tested and prevent regressions in counter accounting.

Minor

Rust test: 20 ms hard-coded timeout

if let Ok(Some(ActorEvent::ConnectionClosed { conn })) =
    tokio::time::timeout(Duration::from_millis(20), events_rx.recv()).await

Testing for the absence of an event necessarily requires a time bound, so some wall-clock wait is unavoidable here. 20 ms is reasonable in practice, but could be marginal on very slow CI workers. Matches the pattern already established in the file for other absence checks, so acceptable as-is.

Rust test: unchecked reply.send return

reply.send(Err(anyhow::anyhow!("reject after transport close")));

The return value is silently discarded. The existing tests in this file follow the same convention (lines 94, 107, 144), so it's consistent — worth noting as a pattern to clean up if the oneshot sender type is ever changed.

Positive notes

• The Rust test cleanly isolates the core invariant: transport close during preflight must never produce a ConnectionClosed event.
• The fixture changes (rejectBefore / rejectCreate) are optional and backward-compatible with no impact on existing tests.
• Using crypto.randomUUID() in the actor key ensures test isolation across runs.
• The test name ("should not call onDisconnect when preflight fails") clearly states the invariant being guarded.