chore: initialize from frontend template 4dc033c

docs/runbooks/FE-RB-001.md

@@ -0,0 +1,9 @@
+# FE-RB-001 — Boot configuration failure
+
+Trigger on `BOOT_CONFIG_FAILURE` after the single bounded refetch fails. Stop
+product route mounting and show the safe support shell; the planned owner
+triage target is five minutes. Escalate from the environment/config owner to
+the release owner.
+
+Close only after a clean-session boot mounts the product root, config
+validation evidence passes, and repeated boot-error telemetry is absent.

docs/runbooks/FE-RB-002.md

@@ -0,0 +1,11 @@
+# FE-RB-002 — Chunk or deployment mismatch
+
+Trigger on `CHUNK_LOAD_FAILURE`, `RELEASE_MANIFEST_FAILURE`, or
+`DEPLOY_MISMATCH`. Warn when dirty state may be lost, fetch the manifest
+`no-store` once, record the release pair, and allow only one reload. The
+planned release-owner triage target is five minutes. Escalate to the hosting/CDN
+owner.
+
+Close only after entry/lazy assets are reachable, the manifest parses into a
+coherent tuple, a second automatic reload is blocked, and the critical route
+smoke passes.

docs/runbooks/FE-RB-003.md

@@ -0,0 +1,10 @@
+# FE-RB-003 — Backend API degradation
+
+Trigger when terminal network/timeout/5xx failures exceed the rolling
+five-minute threshold or on one `SCHEMA_MISMATCH`. Do not expand client retry
+caps, do not retry schema mismatches, and never retry an unkeyed mutation.
+Escalate from the API client owner to backend operations and then release
+compatibility; the planned first-classification target is ten minutes.
+
+Close only after the failure rate returns to baseline, retry amplification is
+absent, critical read/write smoke passes, and schema fixtures pass.

docs/runbooks/FE-RB-004.md

@@ -0,0 +1,9 @@
+# FE-RB-004 — Telemetry sink failure
+
+Trigger on sink network/non-2xx errors, queue overflow, or adapter
+initialization failure. Keep product flows available, bound the queue, and do
+not recursively report to the failed sink. Escalate from observability to the
+telemetry platform owner; the planned triage target is fifteen minutes.
+
+Close only after product e2e remains unaffected, delivery self-check succeeds,
+the queue drains within its bound, and the forbidden-attribute scan passes.

docs/runbooks/FE-RB-005.md

@@ -0,0 +1,13 @@
+# FE-RB-005 — Coherent release rollback
+
+Trigger on a release-blocking boot, chunk, render, API, or security defect when
+a safe forward fix is not demonstrated inside the incident window. Select a
+prior immutable release, verify its asset/config/API tuple, atomically switch
+the complete set, perform the provider cache action, and run smoke checks.
+Escalate from the release-cache owner to the release approver/hosting owner.
+The provider recovery target remains TBD until hosting is selected.
+
+Close only when compatibility and release-coherence gates pass, critical smoke
+passes, repeated `DEPLOY_MISMATCH` is absent, and the incident timeline records
+the restored release ID. Pointer-switch or cache-purge completion alone is not
+recovery evidence.