Multi-process coordinator: per-channel process isolation with auto-restart
breakthroughjobs-applycareerarchitecture
Key Metric
Before
In-process ChannelRunLoop : cross-platform failures cascade
↓
After
Per-channel child processes with WorkerSupervisor (291 li...
Changelog
| Date | Summary |
|---|---|
| 2026-04-07 | Created during temporal gap audit |
| 2026-03-25 | Original work |
Each ATS platform runs in its own isolated child process. WorkerSupervisor (291 lines) manages spawn/crash/auto-restart with exponential backoff (5s/10s/30s/60s), IPC protocol for events, port management, and stale Chrome cleanup. MultiProcessCoordinator (391 lines) aggregates events for Socket.io forwarding, supports per-channel start/stop/pause/resume, staggered launch, and metadata broadcast.
Why It Matters
Replaced the in-process ChannelRunLoop where a crash in one adapter (e.g., LinkedIn restriction) would kill all channels. Process isolation means LinkedIn can crash repeatedly while Direct and Greenhouse continue running. The auto-restart with exponential backoff handles transient failures without manual intervention.