diff --git a/.openclaw-sync/source.json b/.openclaw-sync/source.json
index 30693f5e3..852cc6675 100644
--- a/.openclaw-sync/source.json
+++ b/.openclaw-sync/source.json
@@ -1,5 +1,5 @@
 {
   "repository": "openclaw/openclaw",
-  "sha": "34ef403cb2a21a8013cb4bcef3d319c8b26a31cf",
-  "syncedAt": "2026-04-29T04:03:23.259Z"
+  "sha": "0fc303232506cdcce42306ff71a2b3c657fe2d31",
+  "syncedAt": "2026-04-29T04:27:03.103Z"
 }
diff --git a/docs/ci.md b/docs/ci.md
index 084774f39..7566ead19 100644
--- a/docs/ci.md
+++ b/docs/ci.md
@@ -52,6 +52,14 @@ The native live media shards run in
 live suites on normal Blacksmith runners, because container jobs are the wrong
 place to launch nested Docker tests.
 
+Docker-backed live model/backend shards use a separate shared
+`ghcr.io/openclaw/openclaw-live-test:<sha>` image per selected commit. The live
+release workflow builds and pushes that image once, then the Docker live model,
+gateway, CLI backend, ACP bind, and Codex harness shards run with
+`OPENCLAW_SKIP_DOCKER_BUILD=1`. If those shards rebuild the full source Docker
+target independently, the release run is misconfigured and will waste the wall
+clock on duplicate image builds.
+
 `OpenClaw Release Checks` uses the trusted workflow ref to resolve the selected
 ref once into a `release-package-under-test` tarball, then passes that artifact
 to both the live/E2E release-path Docker workflow and the package acceptance
@@ -216,9 +224,12 @@ manual dispatch; it fans out the mock parity gate, live Matrix lane, and live
 Telegram and Discord lanes as parallel jobs. The live jobs use the
 `qa-live-shared` environment, and Telegram/Discord use Convex leases. Release
 checks run Matrix and Telegram live transport lanes with the deterministic mock
-provider so the channel contract is isolated from live model latency; provider
-connectivity is covered by the separate live model, native provider, and Docker
-provider suites. Matrix uses `--profile fast` for scheduled and release gates,
+provider and mock-qualified models (`mock-openai/gpt-5.5` and
+`mock-openai/gpt-5.5-alt`) so the channel contract is isolated from live model
+latency and normal provider-plugin startup. The live transport gateway also
+disables memory search because QA parity covers memory behavior separately;
+provider connectivity is covered by the separate live model, native provider,
+and Docker provider suites. Matrix uses `--profile fast` for scheduled and release gates,
 adding `--fail-fast` only when the checked-out CLI supports it. The CLI default
 and manual workflow input remain `all`; manual `matrix_profile=all`
 dispatch always shards full Matrix coverage into `transport`, `media`,
diff --git a/docs/help/testing.md b/docs/help/testing.md
index 6d51dd957..150caf950 100644
--- a/docs/help/testing.md
+++ b/docs/help/testing.md
@@ -110,7 +110,17 @@ parallel jobs. Scheduled QA and release checks pass Matrix `--profile fast`
 explicitly, while the Matrix CLI and manual workflow input default remain
 `all`; manual dispatch can shard `all` into `transport`, `media`, `e2ee-smoke`,
 `e2ee-deep`, and `e2ee-cli` jobs. `OpenClaw Release Checks` runs parity plus
-the fast Matrix and Telegram lanes before release approval.
+the fast Matrix and Telegram lanes before release approval, using
+`mock-openai/gpt-5.5` for release transport checks so they stay deterministic
+and avoid normal provider-plugin startup. These live transport gateways disable
+memory search; memory behavior stays covered by the QA parity suites.
+
+Full release live media shards use
+`ghcr.io/openclaw/openclaw-live-media-runner:ubuntu-24.04`, which already has
+`ffmpeg` and `ffprobe`. Docker live model/backend shards use the shared
+`ghcr.io/openclaw/openclaw-live-test:<sha>` image built once per selected
+commit, then pull it with `OPENCLAW_SKIP_DOCKER_BUILD=1` instead of rebuilding
+inside every shard.
 
 - `pnpm openclaw qa suite`
   - Runs repo-backed QA scenarios directly on the host.