release: v4.2.2

CHANGELOG.md

@@ -4,6 +4,10 @@ All notable changes to this project will be documented in this file.
 
 ## [Unreleased]
 
+_Nothing yet._
+
+## [4.2.2] - 2026-06-27
+
 ### Fixed
 - **multi-review**: retry transient reviewer failures instead of killing the review on the first blip. Long-running reviewers (reasoning-effort=max) on self-hosted runners were wiped out by undici `fetch failed` when an upstream stream (opencode server → litellm → model) went quiet past ~300s — observed as 4/6 reviewers dying in the same second. The orchestrator now retries transient network errors (`fetch failed`, ECONNRESET, ETIMEDOUT, UND_ERR_*, …) up to 3 times with exponential backoff, creating a fresh session each attempt to avoid partial-history pollution. Backoff is deadline-aware (gives up instead of sleeping into an immediate re-expiry when the global budget is nearly spent) and jittered (`[base, base+backoffBase]`) so that N reviewers that died together don't all retry together and re-trip the same upstream limit. Our own reviewer deadline timeout is excluded (retrying an expired budget is pointless). Non-transient errors (e.g. `context length exceeded`) are still failed immediately. **Contract change**: a failed attempt now always tears down its session, even under `skipSessionCleanup` (half-run sessions are garbage, not exportable) — only successful sessions are preserved for the v2 export path.
 

multi-review/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-multi-review",
-  "version": "4.2.1",
+  "version": "4.2.2",
   "private": true,
   "type": "module",
   "scripts": {