chore(engine): roxiproxi tests

[NathanFlurry]

Description

Please include a summary of the changes and the related issue. Please also include relevant motivation and context.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

How Has This Been Tested?

Please describe the tests that you ran to verify your changes.

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[NathanFlurry]

• feat(envoy-client): add ping health check #5001
• chore(engine): roxiproxi tests #4864 👈 (View in Graphite)
• fix(rivetkit): preserve query gateway skip ready wait #5008
• fix(depot-client): only fatal sqlite vfs on fence mismatch #5007 : 1 other dependent PR (#5009 )
• fix(rivetkit): preserve internal bridge errors #4993 : 1 other dependent PR (#4997 )
• fix(sqlite): quiet startup database miss #5005
• refactor(metrics): split server metrics crate #5004 : 1 other dependent PR (#5006 )
• fix(rivetkit): use unbounded actor channels #5003
• chore(envoy-client): add envoy key tracing context #5002
• chore(rivetkit): rewrite work registry + fix waituntil not preventing sleep #4990
• main

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

[github-actions]

Preview packages published to npm

Install with:

npm install rivetkit@pr-4864

All packages published as 0.0.0-pr.4864.06929ff with tag pr-4864.

Engine binary is shipped via @rivetkit/engine-cli on linux-x64-musl, linux-arm64-musl, darwin-x64, and darwin-arm64. Windows users should use the release installer or set RIVET_ENGINE_BINARY.

Docker images:

docker pull rivetdev/engine:slim-06929ff
docker pull rivetdev/engine:full-06929ff

Individual packages

npm install rivetkit@pr-4864
npm install @rivetkit/react@pr-4864
npm install @rivetkit/rivetkit-napi@pr-4864
npm install @rivetkit/workflow-engine@pr-4864

[claude]

PR #4864 Review — Network Fault Injection Tests

This PR adds network fault injection tests for the Pegboard Envoy client, verifying that the envoy WebSocket reconnects correctly after a server-side TCP reset without replaying actor start commands. The core architecture is solid, but there are two blocking bugs and several nits.

---

Bug: Duplicate pub mod auth; in envoy/mod.rs

engine/packages/engine/tests/envoy/mod.rs contains pub mod auth; twice. This is a compile error — one of those lines should be pub mod network_faults;:

pub mod auth;
pub mod auth;           // <-- duplicate, will fail to compile
pub mod network_faults;

Bug: on_connect / on_disconnect not wired in connection/wasm.rs

The new callbacks are fired in connection/native.rs but the WASM path (connection/wasm.rs) never calls them. This creates a parity gap — any consumer relying on connection event callbacks will get no events in a WASM build. Per the repo's convention (and CLAUDE.md), both connection paths must be kept at feature parity.

---

Issue: Single proxy port limits ToxiproxyTestServer to one proxy per container

ToxiproxyTestServer::proxy() always listens on the hardcoded DEFAULT_PROXY_PORT (8666). A second proxy created on the same server instance would conflict. The struct name implies multi-proxy support but the implementation doesn't support it. Either document this as a single-proxy limitation or accept a port parameter so the abstraction isn't misleading to future callers.

Nit: wait_for_admin polling loop

toxiproxy.rs uses loop { check; sleep } for the admin API readiness wait. CLAUDE.md flags this pattern, though polling a Docker container's HTTP admin endpoint is a legitimate case with no push notification available. The new wait_with_poll helper introduced in the same PR could be used here for consistency, and a brief comment explaining why event-driven isn't feasible would satisfy the codebase's own guidelines.

Nit: Ordering::SeqCst is stronger than needed

network_faults.rs uses Ordering::SeqCst for the actor start counter. Since there's no ordering relationship established between this counter and other atomics, Ordering::Relaxed is sufficient for both the increment and load.

Nit: ping_actor_via_gateway duplicates ping_actor_via_guard

The local ping_actor_via_gateway function in network_faults.rs appears nearly identical to common::ping_actor_via_guard. If the paths and semantics are equivalent, prefer the shared helper. If they differ intentionally, a comment explaining why is warranted.

---

Required before merge:

1. Remove the duplicate pub mod auth; line — compile error.
2. Add on_connect / on_disconnect to connection/wasm.rs to match native.rs.

The fault injection infrastructure (ToxiproxyTestServer, EnvoyConnectionEventWaiter, wait_with_poll) is well-designed and the test itself covers the right invariant. The nits above can be addressed as follow-up.