[JakeGinnivan]

Upstreamed from dddwa/dddperth#208.

Why this one matters for syncing

When #24 was rebased over #21 (the agenda planner's endless revalidation loop), #24 had moved that logic into the shared useQueuedSave hook, which still had the bug. So the rebase carried #21's edge-triggered fix into the hook. Meanwhile the fork fixed the same loop independently, with a needsRevalidation ref. Left alone, the next /core-pull would conflict on use-queued-save.ts, with two different fixes in the same lines. This PR takes the fork's version, so both sides match.

The five fixes

• admin: revalidate planner data only after queued edits (useQueuedSave, plus a new test)
• agenda: serialize My Agenda shortlist reports, so the first response establishes the HttpOnly identity and add/remove keep their order (use-my-agenda, plus a new test)
• volunteers: reject assignments across overlapping sessions
• sponsors: preserve Jira paragraph and line breaks in rich-text fields
• sponsors: preserve portal saves made during a Jira sync

Upstreamed vs left behind

• All 15 changed files are inside core/; nothing outside it and no baselines
• 14 are byte-identical to the fork's post-#208 copies
• jira-client.test.ts keeps upstream's describe.runIf(conferenceManifest.sponsorPortal) guards (added during the catch-up). #208's new rich-text test builds its own Jira config, so it runs unguarded in the stub.

Testing

• tsc clean; 739 unit tests pass (38 skipped, the same as main, including the 4 guarded Jira tests); eslint clean; build passes
• pnpm vr passes against the existing baselines with none changed, as expected: no rendered output changes. WebKit tablet was re-run after two 35s capture timeouts.
• Did not run the axe e2e suite

After this and #27 merge, the fork's next /core-pull should be close to a no-op for these files. I haven't done a trial pull to confirm it.

🤖 Generated with Claude Code