Pin order 2.10.2 and checkout 0.13.2

[duynhne]

What

The Kind pins for the G4 release: Temporal Go SDK 1.49.0, from #1107.

Workload	Before	After
order (services/order.yaml)	2.10.1	2.10.2
order-worker WorkerDeployment	2.10.1	2.10.2
checkout (services/checkout.yaml)	0.13.1	0.13.2
checkout-worker WorkerDeployment	0.13.1	0.13.2

Release gate

The full local-stack E2E release audit ran from scratch on order d1dafd5 and checkout 8afc665:

• A1–A22 and the A15 versioning drill;
• the 6 k6 suites;
• B1–B10;
• C0–C22.

The first run failed C22: Weaver found the SDK's new failure_reason attribute on temporal_activity_execution_failed (153 violations). duynhlab/pkg#117 declared it in the registry, and the stack was recreated and the whole audit rerun, which passed.

C9 was settled by durable evidence, as the runbook allows: 5 fulfillments Completed = 5 audit orders terminal. The per-process counter reset after the A14/A15 restarts.

The tags carry GitHub Releases, and both images are signed and Trivy-clean.

Verified on Kind (branch overlay, with #1194's controller 1.12 overlaid)

Progressive rollout on controller 1.12, from the controller log:

Time	Worker	Step
14:56:42	order-fulfillment	ramp 10% → 2.10.2-8d45
14:57:12	order-fulfillment	ramp 50%
14:57:43	order-fulfillment	current = 2.10.2-8d45; 2.10.1-fcc8 Draining
14:56:49 / 14:57:19 / 14:57:50	checkout-abandon	same steps → current 0.13.2-f678; 0.13.1-4955 Draining

Status and gates:

• Both WorkerDeployments are ConnectionHealthy, Ready and RolloutComplete.
• make e2e GATE=kind: K4.1–K4.8, K5.2–K5.9 and SG.1–SG.4 all PASS. SG.4 shows Pinned on the new Current build.

Alerts after the rollout, all accounted for:

• KedaScaledObjectErrors, briefly: the new version's ScaledObject queried DescribeWorkerDeploymentVersion at 07:56:42Z, seconds before the version registered. There were no errors after that, and the alert cleared.
• EdgeUpstreamUnhealthy: pending during the pod swap, then cleared.
• TemporalServiceErrorRateHigh, pending at 4.6% (threshold 5%): this is pre-existing. The rule counts NotFound, mostly QueryWorkflow NotFound, which was already at 0.26–0.53/s 3–6 h before this change. A follow-up should exclude NotFound, as the persistence rule already does.
• The firing *HighOverallErrorRate / CheckoutAvailabilityUnknownSKU: produced by the k6 staff/operator suites' deliberate error rows. Yesterday's release recorded the same.

After merge: make flux-push from main. #1194 merges with or before this one, so the controller stays on 1.12.