fix(e2e): bound provision/mint/reap fetches at 45s (hang → fast retry) (#210)

playwright.live.config sets a test timeout but NO per-request timeout, so a
provision (CREATE) or reap (DROP) that intermittently HANGS on postgres-customers
contention runs until the whole test dies (seen as 2-3min vector/cache timeouts).
A hang isn't slowness — raising the test timeout just makes it last longer. Bound
the provision POST, the mint, and both reap DELETEs at 45s so a hang fails fast
and Playwright retries (passes when contention clears). Pairs with the
orphan-customer-DB sweep that removes the contention source.

Co-authored-by: Claude <noreply@anthropic.com>

Author: mastermanas805

e2e/cleanup-ledger.ts

@@ -184,6 +184,10 @@ export async function reapEntities(
         method: 'DELETE',
         headers,
         failOnStatusCode: false,
+        // Bound the reap DELETE: a DROP that hangs on postgres-customers
+        // contention must fail fast (45s) rather than hang teardown to the test
+        // timeout. The on-disk ledger backstop re-reaps on the next run.
+        timeout: 45_000,
       })
       const status = resp.status()
       const bodyText = status >= 200 && status < 300 ? '' : await resp.text().catch(() => '<unreadable>')

e2e/factory.ts

@@ -135,6 +135,10 @@ export async function mintUser(
       with_failed_deploy: !!opts.withFailedDeploy,
     }),
     failOnStatusCode: false,
+    // Bound the mint: with_resources provisions real DBs server-side, which can
+    // hang on postgres-customers contention. 45s fails fast → the spec SKIPs
+    // (mint==null) rather than hanging the whole test on a stalled mint.
+    timeout: 45_000,
   })
   // Inert-by-default 404: the token is wrong or the endpoint isn't armed on
   // this stack. Treat as "can't mint" → null (caller SKIPS), never a red.
@@ -217,6 +221,9 @@ export async function reap(request: APIRequestContext, teamID: string): Promise<
     method: 'DELETE',
     headers: { [E2E_ACCOUNT_TOKEN_HEADER]: accountToken() },
     failOnStatusCode: false,
+    // Bound the cohort reap: the team-cascade drops customer DBs; a hung DROP
+    // must fail fast (45s), not hang teardown. afterAll + ledger reaper backstop.
+    timeout: 45_000,
   })
   const ok = [200, 202, 204, 404, 410].includes(resp.status())
   if (!ok) {

e2e/live-anon-provision.spec.ts

@@ -281,6 +281,12 @@ test.describe('LIVE — every anonymous provision flow → backend-assert → re
         headers: id.headers,
         data: JSON.stringify({ name }),
         failOnStatusCode: false,
+        // Bound the provision call: a CREATE that intermittently hangs on
+        // postgres-customers contention would otherwise eat the whole test
+        // timeout (a hang, not slowness — raising the test timeout just makes it
+        // last longer). 45s fails fast → Playwright retries → passes on the next
+        // attempt when the contention clears. Pairs with the orphan-DB sweep.
+        timeout: 45_000,
       })
 
       test.skip(