Fix giscus like button not persisting on first reaction

[soyalejolopez]

Problem

Clicking "like" (giscus reaction) on a resource that has no backing Discussion yet makes the 👍 vanish instead of sticking. This is giscus bug ##1312: the first reaction creates the discussion + reaction server-side, but the giscus client never refetches, so the reaction visually reverts.

Fix

• Event-driven remount: track existedAtMount via the giscus emit-metadata listener and force a single refetch remount once a discussion is created by a first reaction.
• Removed the self-defeating hasDiscussion re-check from the blur-timer callback that was cancelling the very remount meant to persist the like.
• Added a speculativeRefetch flag + re-arm branch so focusing the iframe without reacting (e.g. to sign in) doesn't permanently block recovery.

Validation

• Local Playwright load: zero console errors, giscus mounts, detail view renders.
• Peer-reviewed by GPT 5.6 (caught an edge case, re-reviewed PASS).

Ref: giscus/giscus#1312

[github-actions]

🛰️ PR Sweeper report

Risk: 🟢 LOW · Security gate: ✅ passing · Files: 1

🔒 Automated guardrails (authoritative)

	Category	Location	Finding
ℹ️	scope	—	Change(s) outside known content roots: index.html.

🤖 Dual-model AI review (advisory)

Deep-reasoning reviewer — openai/o3 · verdict: approve · risk: low

PR adjusts giscus discussion-recovery logic, adds two state flags and slightly increases the blur fallback timeout. No secrets, dangerous calls, or privacy exposures detected. Change is self-contained and aligns with existing style.

Quality notes:

• index.html — Clear inline comments improve code readability; uses non-minified JS with no new external dependencies. No repo metadata changes needed for this small bug-fix style update.

✅ Suggested next steps

• Consider adding a short comment in the commit message or PR description explaining the rationale for the 3 s timeout to help future maintainers.
• If automated tests exist for comment-thread behavior, add a test case covering speculative refetch re-arming.

---

The automated guardrails are authoritative and gate the security status. The AI review is advisory and never auto-merges. Thanks for contributing to FastTrack! 🛩️