docs(release): prepare v1.1.26

Author: PiyushDatta

release_log.md

@@ -2,18 +2,18 @@
 
 ## Release Metadata
 
-- version: `v1.1.25`
-- start_commit: `d6ffb6d1c1112be2d0a6bbac079895d22b5ec564`
-- end_commit: `585b57a0ba94545fdcba27f3cfed32acb9669b84`
-- commits_in_range: `4`
+- version: `v1.1.26`
+- start_commit: `e1a5ba4910e4020590a6386d638071eda62d0dd9`
+- end_commit: `2b586c0976968d3f064ff6f5e58ace21e19c404f`
+- commits_in_range: `2`
 
 ## Highlights
 
-- Shorten no-edit recovery for stuck OpenAI Codex WorkerPal jobs so off-track tasks fail or recover faster instead of burning a full execution window.
-- Skip low-value critic retries after deterministic fast validation failures, preserving remaining job budget for actionable repair attempts.
-- Reduce duplicate cooldowns for narrow contract/test retry families so WorkerPals can respond to active validation noise without waiting through stale long cooldowns.
-- Sanitize stale planner path hints before WorkerPal execution, including stale prose guidance, so workers do not chase nonexistent repo prefixes such as outdated client/app paths.
-- Ship the updated WorkerPal sandbox mirror in the packaged CLI runtime assets.
+- Add an autonomy WorkerPal failure circuit for repeated no-publishable/no-edit outcomes so background autonomy stops creating new jobs during unhealthy windows.
+- Suppress repeated autonomy jobs that match recent no-publishable failures by pattern key or overlapping target paths, with structured retry metadata instead of another long WorkerPal attempt.
+- Teach RemoteBuddy autonomy dispatch to honor structured enqueue rejections with a timed backoff, reducing repeat dispatch loops after safety gates fire.
+- Add bounded WorkerPal cooldown metadata for OpenAI Codex no-publishable and broad/off-track timeout failures so retry storms pause after exhausted attempts.
+- Ship the updated RemoteBuddy and WorkerPal sandbox runtime mirror in the packaged CLI assets.
 
 ## Validation
 
@@ -46,7 +46,7 @@ bun install -g @pushpalsdev/cli
 ## Known Issues
 
 - Docker-backed WorkerPal execution still requires Docker to be installed and running when WorkerPal auto-spawn is enabled; `pushpals --clear` cleanup is best-effort when Docker is unavailable or times out.
-- Active runtimes that were started from `v1.1.24` or earlier must be restarted after installing this release before stale-path sanitization and faster WorkerPal recovery behavior take effect.
+- Active runtimes that were started from `v1.1.25` or earlier must be restarted after installing this release before the new autonomy failure circuits, dispatch backoff, and WorkerPal cooldown behavior take effect.
 - Codex `gpt-5.5` requires a recent Codex CLI; older Codex CLIs fall back to `gpt-5.4` for WorkerPal and RemoteBuddy Codex execution when they report model incompatibility.
 - GitHub contribution credit for WorkerPal commits requires the configured commit email to be associated with the target GitHub account.
 - User-local `runtime/configs/local.toml` overrides can preserve older runtime defaults during manual smoke testing; use `pushpals --clear` or remove the local override to pick up new packaged defaults.