Remote comms: Fix remaining TODO tests

[sirtimid]

Overview

Most failure tests for remote communications have been implemented. This issue tracks the remaining TODO tests that need to be fixed or completed.

Completed ✅

All remaining TODO tests have been fixed.

1. Transport layer race condition test

File: packages/ocap-kernel/src/remotes/platform/transport.test.ts:1219

Implemented in PR #788 (kernel incarnation detection protocol).

2. E2E connection failure recovery test

File: packages/nodejs/test/e2e/remote-comms.test.ts:267

Fixed by restructuring the test to queue the recovery message while the peer is down (triggering reconnection logic) rather than after the peer restarts.

Already Implemented ✅

The following failure scenarios are covered by existing tests:

Unit Tests (~3,500+ lines)

• transport.test.ts: Rate limiting, connection limits, reconnection, stale peer cleanup
• reconnection-lifecycle.test.ts: Backoff calculations, retry exhaustion, non-retryable errors
• RemoteHandle.test.ts: ACK protocol, retransmission, crash recovery, queue limits

E2E Tests (~900 lines in remote-comms.test.ts)

• Kernel restart survival
• Reconnection with exponential backoff
• Connection to non-existent peer (graceful failure)
• Message queueing when offline
• Queue limit enforcement (MAX_QUEUE = 200)
• Multiple simultaneous peer reconnections
• Intentional disconnect handling
• Promise rejection on remote give-up
• Promise resolution after reconnection
• Connection failure and recovery (fixed)

Acceptance Criteria

• Transport race condition test rewritten for ACK protocol (PR feat(remote-comms): Add kernel incarnation detection protocol #788)
• E2E connection recovery test fixed and passing