refactor(iscored): one SWR cache helper and safeFetch for both fetches (PP-az4d.19)

[timothyfroehlich]

Summary

• Replace the two hand-rolled in-memory stale-while-revalidate caches in src/lib/iscored/client.ts with one createSwrCache<T>({ttlMs, failureRetryMs, load, empty}) helper. The genuine difference stays: a failed games fetch retries after 15s, a failed scores fetch waits the normal 15s TTL.
• Move both raw fetch calls onto safeFetch from ~/lib/http/external (CORE-ARCH-016) behind one fetchIscoredJson, keeping the 8s timeout, log messages, and result shapes.
• Add two tests pinning the retry-interval difference; note iScored's move in CORE-ARCH-016.

No behavior change. Tests: src/lib/iscored unit (57 pass), iscored-actions integration (13 pass), pnpm run check clean.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
pin-point	Ready	Preview	Oct 6, 2026 7:53pm UTC

[timothyfroehlich]

Claude Code review (medium)

Reviewed head 9a2e558 with /code-review medium. 2 findings: 1 fixed, 1 declined.

Round	Finding	Disposition
1	src/lib/iscored/client.ts:175 — Mid-body read failure now goes to reportError instead of log.warn and could raise a Sentry event per refresh during an iScored outage.	Fixed in 9a2e558
2	src/lib/iscored/client.ts:78 — An in-flight load for an old gameroom user can store its result under a new key after a user change or clear().	Declined: Pre-existing behavior carried over unchanged by this no-behavior-change refactor, and ISCORED_USER is a fixed per-process env value so the key never changes at runtime.

—Claude