ci: workarounds for scale to zero by AlexAxthelm · Pull Request #238 · RMI/stitch

AlexAxthelm · 2026-08-17T16:46:36Z

Alex's TL;DR here, Claude's fuller summary below the fold:

Last pinged around 1510 Belgrade time. Update this description as needed.

Responding to changes in #234, where we set more environments to scale to zero (fall asleep) this PR is a collection of workarounds to mitigate the impact of that in practice:

(up for discussion after we try it out): Set timeout before falling asleep to 15 min (900s) up from 5 min (300s)
Add "waking up" message on env banner (this only affects the same environments as get env banner; prod is always-on). See screenshot below
ping the API as part of the initial page load (prior to login) so server can be spinning up while user is (maybe) running through login
add utility to frontend to help it understand which erros should be retried.

Why

#234 narrows the always-on policy so that only pushes to production keep a warm replica. Everything else — main, demo lanes, PR previews — now scales to zero and pays a cold start on the first request of a session. This PR makes that acceptable: slightly less frequent, and no longer indistinguishable from the app being broken.

Widening the cooldown

Cooldown is how long a Container App stays up after its last request. Azure's default is 300s, so a five-minute pause costs a cold start. This raises it to 900s for apps that actually scale to zero, reasserted after every deploy.

Two things a reviewer should know about it:

There is no CLI flag. az containerapp update has no --cooldown-period and no generic --set, only --yaml. The step therefore does the same show → jq → update --yaml round-trip already used for ETL volume mounts, and runs after that step so it re-reads its result. The jq transform was dry-run against the live main-api spec: cooldownPeriod 300 → 900, systemData stripped (ARM rejects it on write), image digest and the rest of the template intact.

Absorbing the cold start in the browser

Retry transient failures instead of surfacing them. The global retry: 1 retried once after ~1s — still inside an ACA activation — then showed the user an error. Now a predicate retries only 5xx, 408, 429 and the TypeError a dropped fetch rejects with, up to 3 times.

Wake the API during bootstrap. A throwaway GET fired from bootstrap() and not awaited. It runs before Auth0 mounts, so the container starts while the user is still being redirected through login. Targets /health/details rather than /health because it resolves the database engine, so it opens a connection too — the two costs of a cold start, both paid during dead time. Neither endpoint requires auth.

Say so when it is still slow. After 2s of a request in flight, the environment banner shows "Server is waking up — this can take a few seconds" as a role="status" live region. Nothing to fix in that moment; it just needs saying, so the wait reads as starting up rather than nothing happening.

Details that look like omissions but are deliberate

Each is commented at the call site and pinned by a test, so they are not tidied away later.

Choice	Reason
The cooldown step is gated on `min-replicas == ''`	An app that never scales to zero has no use for a cooldown, and skipping it saves a round-trip per deploy
Prewarm has no abort timer	Azure holds the request open while it activates a replica — that wait is the thing doing the work
Prewarm sends no headers and no auth	Keeps it CORS-simple, so it reaches the server without a preflight even when the deployment's single allowed origin does not match the browser's. The response is discarded, so being blocked from reading it costs nothing — which matters during the `stitch-dev.rmi.org` cutover
Prewarm never rejects	A failed prewarm has cost the user nothing; the real request behind it carries its own retry policy
The warming hook keys on a boolean, not the fetch count	Otherwise a second query starting restarts the 2s clock on the request that was already waiting
Mutations keep `retry: 0`	They are not idempotent; replaying a write after a timeout is worse than the error

Things to look at

A 4xx is no longer retried. retry: 1 retried every failure once, including 404 and 401. Re-sending an unchanged bad request only delays an error that was never going to change — but this is a change beyond "retry more", so it is worth a reviewer's eye.

getResources and getResourceFilterOptions now attach error.status. Every other throw site already did; these two carried the status only in the message string, so a status-based retry predicate could not see them. They back the resources list and its filter options — the first requests after sign-in, and so exactly the ones that meet a cold container. The retry fix does not work without this.

The cooldown step costs CI time. A show plus an update per scale-to-zero app per deploy — four apps on a PR lane, so perhaps 30–60s. Cheap, but not free, and easy to forget once buried in a reusable workflow.

Existing apps stay at 300s until their next deploy.

Not in this PR

Prewarming entity-linkage and stitch-llm. Built, then reverted — better approaches are being explored, and the API is the one every session needs.
An HTTP startup probe on /health. No probes are configured on any container app, so ACA falls back to TCP-open, and uvicorn can accept connections before the FastAPI lifespan finishes. That is a plausible source of the 503s the retry policy is defending against. It needs the same --yaml round-trip, so it is CI work and belongs in its own PR.
Tuning retryDelay. Left at the TanStack default (1s/2s/4s). It is the knob to turn if the retries prove too eager against a real cold start, but picking a curve now would be guessing.

Notes for review

Stacked on #234 — base is ci/hostnames, so that needs to merge first.

AI was used to write the frontend changes, the CI step, the tests and this description. The measurements are from az monitor metrics list against main-api, not estimates. The new banner tests were mutation-checked: forcing the delay never to elapse kills exactly the two tests asserting the notice appears and clears, and leaves the other six green.

github-actions · 2026-08-17T16:48:41Z

CD summary `6510986`

Database (1)

db_name	postgres_host	postgres_port	postgres_db
pr_0238	`stitch-dev.postgres.database.azure.com`	`5432`	`pr_0238`

Jobs (1)

job	image	postgres_db
db-migrations	`ghcr.io/rmi/stitch-api:pr-0238@sha256:652799578d8e819f128c2dc72fc4bf5e1414ad1f77340296893a0490ae86b7e6`	`pr_0238`

Images (4)

build_time	commit_time	git_sha	image	image_digest
2026-08-17T16:46:57Z	2026-08-17T16:46:36Z	`c20e793`	`ghcr.io/rmi/stitch-api:pr-0238`	`ghcr.io/rmi/stitch-api:pr-0238@sha256:652799578d8e819f128c2dc72fc4bf5e1414ad1f77340296893a0490ae86b7e6`
2026-08-17T16:46:57Z	2026-08-17T16:46:36Z	`c20e793`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238@sha256:e2263ae5a66f6b319526391f83f71fe4d7a904a35be4a0c450eff43414a9d3df`
2026-08-17T16:46:55Z	2026-08-17T16:46:36Z	`c20e793`	`ghcr.io/rmi/stitch-seed:pr-0238`	`ghcr.io/rmi/stitch-seed:pr-0238@sha256:0c78a5d335d1a44b30f471def0f0b2193f6bebda7e4e6880b14f980328d0d6e9`
2026-08-17T16:46:56Z	2026-08-17T16:46:36Z	`c20e793`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238@sha256:7ae51cc8127f51511af7438996c089b4fe84676a45407c3a0c996be1c6073f6a`

github-actions · 2026-08-17T16:52:56Z

CD summary `9556f57`

Frontend: https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net

Deployments (4)

service	url	fqdn
api	open	`pr-0238-api.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
entity-linkage	open	`pr-0238-el.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
frontend	https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net
stitch-llm	open	`pr-0238-llm.purplegrass-c07d0a94.westus2.azurecontainerapps.io`

Database (1)

db_name	postgres_host	postgres_port	postgres_db
pr_0238	`stitch-dev.postgres.database.azure.com`	`5432`	`pr_0238`

Jobs (1)

job	image	postgres_db
db-migrations	`ghcr.io/rmi/stitch-api:pr-0238@sha256:8064a9e16c4a562914933878eaff4a38ffbad5b07f79128f14d32500e22ac884`	`pr_0238`

Images (4)

build_time	commit_time	git_sha	image	image_digest
2026-08-17T16:49:00Z	2026-08-17T16:48:25Z	`ad14cdb`	`ghcr.io/rmi/stitch-api:pr-0238`	`ghcr.io/rmi/stitch-api:pr-0238@sha256:8064a9e16c4a562914933878eaff4a38ffbad5b07f79128f14d32500e22ac884`
2026-08-17T16:49:03Z	2026-08-17T16:48:25Z	`ad14cdb`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238@sha256:b0531bf3b6d702b20d1f646367cf38bfaa41afc550933353c530f83812d99bd8`
2026-08-17T16:49:02Z	2026-08-17T16:48:25Z	`ad14cdb`	`ghcr.io/rmi/stitch-seed:pr-0238`	`ghcr.io/rmi/stitch-seed:pr-0238@sha256:c15480b89ce79515ccf22e2907d01e02e046eab03e24c5ac7c6ca46841683f0a`
2026-08-17T16:48:57Z	2026-08-17T16:48:25Z	`ad14cdb`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238@sha256:546cc49b3c9d23db85854dd7b2408f7215ca63065b0553e3fb146b16b4ebdfcc`

github-actions · 2026-08-18T16:56:06Z

CD summary `f6daee1`

Frontend: https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net

Deployments (4)

service	url	fqdn
api	open	`pr-0238-api.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
entity-linkage	open	`pr-0238-el.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
frontend	https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net
stitch-llm	open	`pr-0238-llm.purplegrass-c07d0a94.westus2.azurecontainerapps.io`

Database (1)

db_name	postgres_host	postgres_port	postgres_db
pr_0238	`stitch-dev.postgres.database.azure.com`	`5432`	`pr_0238`

Jobs (1)

job	image	postgres_db
db-migrations	`ghcr.io/rmi/stitch-api:pr-0238@sha256:340f53c0a867908e79a15173f8b6c7b2d680aac6ade02530db89228ad5ebacb5`	`pr_0238`

Images (4)

build_time	commit_time	git_sha	image	image_digest
2026-08-18T16:51:04Z	2026-08-18T16:50:47Z	`6de44b9`	`ghcr.io/rmi/stitch-api:pr-0238`	`ghcr.io/rmi/stitch-api:pr-0238@sha256:340f53c0a867908e79a15173f8b6c7b2d680aac6ade02530db89228ad5ebacb5`
2026-08-18T16:51:00Z	2026-08-18T16:50:47Z	`6de44b9`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238@sha256:aac380f27bb220fd9d93c5b6441469e57652099daf0a4dbef5aead440b58f044`
2026-08-18T16:51:00Z	2026-08-18T16:50:47Z	`6de44b9`	`ghcr.io/rmi/stitch-seed:pr-0238`	`ghcr.io/rmi/stitch-seed:pr-0238@sha256:d4a285f9915b8c460b40ed2e3791fa90fbf49f7d96d47fe47188d621391836a4`
2026-08-18T16:51:02Z	2026-08-18T16:50:47Z	`6de44b9`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238@sha256:0d7f3f47f1741eda61c31333aa593df92c01ba258cc9649aadac688bf92db5a0`

github-actions · 2026-08-20T11:36:36Z

CD summary `de1a696`

Frontend: https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net

Deployments (4)

service	url	fqdn
api	open	`pr-0238-api.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
entity-linkage	open	`pr-0238-el.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
frontend	https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net
stitch-llm	open	`pr-0238-llm.purplegrass-c07d0a94.westus2.azurecontainerapps.io`

Database (1)

db_name	postgres_host	postgres_port	postgres_db
pr_0238	`stitch-dev.postgres.database.azure.com`	`5432`	`pr_0238`

Jobs (1)

job	image	postgres_db
db-migrations	`ghcr.io/rmi/stitch-api:pr-0238@sha256:a9d1c1e9f236d77e9a19c88a473f2aab32854b72f58cef89fe5c1e9132acb8b6`	`pr_0238`

Images (4)

build_time	commit_time	git_sha	image	image_digest
2026-08-20T11:31:26Z	2026-08-20T11:30:56Z	`32a5be5`	`ghcr.io/rmi/stitch-api:pr-0238`	`ghcr.io/rmi/stitch-api:pr-0238@sha256:a9d1c1e9f236d77e9a19c88a473f2aab32854b72f58cef89fe5c1e9132acb8b6`
2026-08-20T11:31:13Z	2026-08-20T11:30:56Z	`32a5be5`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238@sha256:52d782b88bff07e7153db9dffe59b2330c5181d90f0b1454cfc552fdc540a85c`
2026-08-20T11:31:13Z	2026-08-20T11:30:56Z	`32a5be5`	`ghcr.io/rmi/stitch-seed:pr-0238`	`ghcr.io/rmi/stitch-seed:pr-0238@sha256:8caeeb742504b5736bc0266c549fbc2d0f1af6412128f6f9cae56571b144b0cb`
2026-08-20T11:31:11Z	2026-08-20T11:30:56Z	`32a5be5`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238@sha256:2057afd2acc3130f5a813ed4d43cb40396cd19ef7268a3e6397015b24c825fe5`

github-actions · 2026-08-20T13:21:07Z

CD summary `de1a696`

Frontend: https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net

Deployments (4)

service	url	fqdn
api	open	`pr-0238-api.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
entity-linkage	open	`pr-0238-el.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
frontend	https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net
stitch-llm	open	`pr-0238-llm.purplegrass-c07d0a94.westus2.azurecontainerapps.io`

Database (1)

db_name	postgres_host	postgres_port	postgres_db
pr_0238	`stitch-dev.postgres.database.azure.com`	`5432`	`pr_0238`

Jobs (2)

job	image	postgres_db	api_url	auth_mode
db-migrations	`ghcr.io/rmi/stitch-api:pr-0238@sha256:d616dfd54d329f9d3bbc40dfdee7830ca910244b1e1481a221cdaa246a01fb0d`	`pr_0238`
seed	`ghcr.io/rmi/stitch-seed:pr-0238@sha256:144be6d630fdee428763a90bb45978b617fe8642d766bcdc4e7832daf0e8103c`		`https://pr-0238-api.purplegrass-c07d0a94.westus2.azurecontainerapps.io/api/v1`	`stitch-client-bearer-token`

Images (4)

build_time	commit_time	git_sha	image	image_digest
2026-08-20T13:16:05Z	2026-08-20T13:15:36Z	`32a5be5`	`ghcr.io/rmi/stitch-api:pr-0238`	`ghcr.io/rmi/stitch-api:pr-0238@sha256:d616dfd54d329f9d3bbc40dfdee7830ca910244b1e1481a221cdaa246a01fb0d`
2026-08-20T13:15:51Z	2026-08-20T13:15:36Z	`32a5be5`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238@sha256:31efc37cc73e8a182bdb23a4630bc9e6ce8d39715739b7c115f81311a33391ae`
2026-08-20T13:15:53Z	2026-08-20T13:15:36Z	`32a5be5`	`ghcr.io/rmi/stitch-seed:pr-0238`	`ghcr.io/rmi/stitch-seed:pr-0238@sha256:144be6d630fdee428763a90bb45978b617fe8642d766bcdc4e7832daf0e8103c`
2026-08-20T13:15:49Z	2026-08-20T13:15:36Z	`32a5be5`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238@sha256:a3a82adf319f5c8838b02896d2c4eda05975b5eac17b36582dbe0beae5983ab5`

Copilot

Pull request overview

This PR mitigates the user-facing impact of Azure Container Apps scaling to zero (cold starts) by (1) widening the cooldown period in CI/CD and (2) making the frontend proactively “wake” the API and retry only genuinely transient failures, while also surfacing an explicit “server is waking up” message during stalled requests.

Changes:

CI: add a post-deploy step to set cooldownPeriod to 900s for scale-to-zero Container Apps via show → jq → update --yaml.
Frontend: prewarm the API during bootstrap and introduce a TanStack Query retry predicate that retries only transient failures (5xx/408/429 + failed fetch).
UI/tests: show an accessible “Server is waking up” notice after a short delay while requests are pending, with accompanying tests.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
`.github/workflows/deploy-container.yml`	Adds a deploy step to widen ACA scale-to-zero cooldown to reduce cold starts within a session.
`deployments/CI_DEPLOYMENTS.md`	Documents the widened cooldown and the frontend mitigations; adds a metrics command to measure impact.
`deployments/stitch-frontend/src/AppProviders.jsx`	Switches TanStack Query `retry` from a fixed count to a predicate for transient-only retries.
`deployments/stitch-frontend/src/main.jsx`	Fires a non-blocking API prewarm call during bootstrap so warm-up overlaps with Auth0 redirect time.
`deployments/stitch-frontend/src/queries/prewarm.js`	Implements a best-effort prewarm request to `/health/details` that should never block bootstrap.
`deployments/stitch-frontend/src/queries/prewarm.test.js`	Adds unit tests for prewarm behavior (URL, CORS-simplicity, non-blocking semantics).
`deployments/stitch-frontend/src/queries/retry.js`	Adds `isRetriableError` and `shouldRetryQuery` utilities to classify retryable failures.
`deployments/stitch-frontend/src/queries/retry.test.js`	Adds unit tests for retry classification and retry cap behavior.
`deployments/stitch-frontend/src/queries/api.js`	Ensures thrown HTTP errors include `error.status` so retry logic can inspect status codes.
`deployments/stitch-frontend/src/components/EnvironmentBanner.jsx`	Adds a delayed “waking up” live region message when queries appear stalled.
`deployments/stitch-frontend/src/components/EnvironmentBanner.test.jsx`	Adds tests for the “waking up” notice timing and clearing behavior.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

+export function prewarmApi(config) {
+  const baseUrl = config?.apiBaseUrl;
+
+  if (!baseUrl) {
+    return Promise.resolve();
+  }
+
+  // Deliberately not aborted on a timer: Azure holds the request open while it
+  // activates a replica, and that wait is the thing doing the work.
+  //
+  // Deliberately unauthenticated and header-free, which keeps this a
+  // CORS-simple request. It reaches the server without a preflight even when
+  // the deployment's single allowed origin does not match the browser's -- and
+  // since the response is discarded, being blocked from reading it costs
+  // nothing. That matters while a custom domain is being cut over.
+  return fetch(healthUrl(baseUrl), { cache: "no-store" }).then(
+    () => undefined,
+    () => undefined,
+  );
+}


+import { describe, it, expect, vi, beforeEach } from "vitest";
+import { prewarmApi } from "./prewarm";
+
+describe("prewarmApi", () => {
+  let fetchMock;
+
+  beforeEach(() => {
+    fetchMock = vi.fn().mockResolvedValue({ ok: true });
+    vi.stubGlobal("fetch", fetchMock);
+  });
+


+lane is in real daily use:
+
+```bash
+az monitor metrics list --resource "$(az containerapp show -n main-api -g STITCH-DEV-RG --query id -o tsv)" --metric Replicas --interval PT1H --start-time "$(date -u -v-13d +%Y-%m-%dT%H:%M:%SZ)" --end-time "$(date -u +%Y-%m-%dT%H:%M:%SZ)" --aggregation Maximum -o table


* Guard `prewarmApi` against a non-string base URL. Unreachable via `loadConfig`, which already rejects one, but the function documents that it never rejects and `main.jsx` relies on that -- a synchronous throw would escape `bootstrap()` and swap the app for the config-error screen. * Unstub `fetch` in `afterEach`, matching the other suites that stub globals. `vitest.config.js` sets no `unstubGlobals`, and the shared `restoreAllMocks` does not undo `stubGlobal`. * Replace the BSD-only `date -v-13d` in the metrics snippet with `az monitor metrics list --offset 13d`, which needs no date arithmetic and works on GNU too. Phrase the sample as a ratio rather than an exact hour count, which drifts as the window moves. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

github-actions · 2026-08-20T17:20:51Z

CD summary `49effe8`

Frontend: https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net

Deployments (4)

service	url	fqdn
api	open	`pr-0238-api.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
entity-linkage	open	`pr-0238-el.purplegrass-c07d0a94.westus2.azurecontainerapps.io`
frontend	https://witty-mushroom-017a3dc1e-238.westus2.1.azurestaticapps.net
stitch-llm	open	`pr-0238-llm.purplegrass-c07d0a94.westus2.azurecontainerapps.io`

Database (1)

db_name	postgres_host	postgres_port	postgres_db
pr_0238	`stitch-dev.postgres.database.azure.com`	`5432`	`pr_0238`

Jobs (1)

job	image	postgres_db
db-migrations	`ghcr.io/rmi/stitch-api:pr-0238@sha256:2fef95c5931453385318a21fef949768b8c3482a293edf860f798ab64acee528`	`pr_0238`

Images (4)

build_time	commit_time	git_sha	image	image_digest
2026-08-20T17:15:41Z	2026-08-20T17:15:22Z	`77bfdae`	`ghcr.io/rmi/stitch-api:pr-0238`	`ghcr.io/rmi/stitch-api:pr-0238@sha256:2fef95c5931453385318a21fef949768b8c3482a293edf860f798ab64acee528`
2026-08-20T17:15:40Z	2026-08-20T17:15:22Z	`77bfdae`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238`	`ghcr.io/rmi/stitch-entity-linkage:pr-0238@sha256:a614c1d6491d8693105e4f6853140840520fbd67675adc3920dd8f148536105c`
2026-08-20T17:15:44Z	2026-08-20T17:15:22Z	`77bfdae`	`ghcr.io/rmi/stitch-seed:pr-0238`	`ghcr.io/rmi/stitch-seed:pr-0238@sha256:2dc781685f1ea429c23d5af4e04fe551804ab0be8a4a6382b864b80261404adc`
2026-08-20T17:15:40Z	2026-08-20T17:15:22Z	`77bfdae`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238`	`ghcr.io/rmi/stitch-stitch-llm:pr-0238@sha256:1e77b6b24e14440ac48dcb2513ab08bd75797ee28a0d3bdda80802c066754b7e`

AlexAxthelm deployed to development August 17, 2026 16:46 — with GitHub Actions Active

AlexAxthelm deployed to development August 17, 2026 16:47 — with GitHub Actions Active

AlexAxthelm had a problem deploying to development August 17, 2026 16:48 — with GitHub Actions Error

AlexAxthelm deployed to development August 17, 2026 16:48 — with GitHub Actions Active

AlexAxthelm deployed to development August 17, 2026 16:49 — with GitHub Actions Active

AlexAxthelm deployed to development August 17, 2026 16:50 — with GitHub Actions Active

AlexAxthelm deployed to development August 17, 2026 16:51 — with GitHub Actions Active

AlexAxthelm deployed to development August 18, 2026 16:50 — with GitHub Actions Active

AlexAxthelm deployed to development August 18, 2026 16:52 — with GitHub Actions Active

AlexAxthelm deployed to development August 18, 2026 16:54 — with GitHub Actions Active

AlexAxthelm deployed to development August 18, 2026 16:55 — with GitHub Actions Active

AlexAxthelm deployed to development August 18, 2026 17:28 — with GitHub Actions Active

AlexAxthelm deployed to development August 18, 2026 17:29 — with GitHub Actions Active

AlexAxthelm deployed to development August 18, 2026 17:31 — with GitHub Actions Active

AlexAxthelm deployed to development August 20, 2026 11:33 — with GitHub Actions Active

AlexAxthelm deployed to development August 20, 2026 11:35 — with GitHub Actions Active

AlexAxthelm added the no-deploy label Aug 20, 2026

AlexAxthelm deployed to development August 20, 2026 13:08 — with GitHub Actions Active

AlexAxthelm removed the no-deploy label Aug 20, 2026

AlexAxthelm deployed to development August 20, 2026 13:15 — with GitHub Actions Active

AlexAxthelm deployed to development August 20, 2026 13:16 — with GitHub Actions Active

AlexAxthelm deployed to development August 20, 2026 13:17 — with GitHub Actions Active

AlexAxthelm deployed to development August 20, 2026 13:18 — with GitHub Actions Active

AlexAxthelm deployed to development August 20, 2026 13:19 — with GitHub Actions Active

AlexAxthelm marked this pull request as ready for review August 20, 2026 13:21

Copilot AI lite review requested due to automatic review settings August 20, 2026 13:21

Copilot started reviewing on behalf of AlexAxthelm August 20, 2026 13:22 View session

AlexAxthelm requested review from john and mbarlow12 August 20, 2026 13:22

Copilot AI reviewed Aug 20, 2026

View reviewed changes

Navigation Menu

Uh oh!

ci: workarounds for scale to zero - #238

ci: workarounds for scale to zero#238
AlexAxthelm wants to merge 7 commits into
mainfrom
ci/cooldown

AlexAxthelm commented Aug 17, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 17, 2026

Uh oh!

github-actions Bot commented Aug 17, 2026

Uh oh!

github-actions Bot commented Aug 18, 2026

Uh oh!

github-actions Bot commented Aug 20, 2026

Uh oh!

github-actions Bot commented Aug 20, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

github-actions Bot commented Aug 20, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

AlexAxthelm commented Aug 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Why

Widening the cooldown

Absorbing the cold start in the browser

Details that look like omissions but are deliberate

Things to look at

Not in this PR

Notes for review

Uh oh!

github-actions Bot commented Aug 17, 2026

CD summary 6510986

Uh oh!

github-actions Bot commented Aug 17, 2026

CD summary 9556f57

Uh oh!

github-actions Bot commented Aug 18, 2026

CD summary f6daee1

Uh oh!

github-actions Bot commented Aug 20, 2026

CD summary de1a696

Uh oh!

github-actions Bot commented Aug 20, 2026

CD summary de1a696

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

github-actions Bot commented Aug 20, 2026

CD summary 49effe8

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

AlexAxthelm commented Aug 17, 2026 •

edited

Loading

CD summary `6510986`

CD summary `9556f57`

CD summary `f6daee1`

CD summary `de1a696`

CD summary `de1a696`

CD summary `49effe8`