fix(board): boot once — stop the issue-list flicker on mount (v0.1.3)

[mabry1985]

Closes #13.

Root cause

Both view pages called boot() twice:

kit.initPluginView(boot);   // the kit invokes boot() after the theme/auth handshake
boot();                     // …and a direct call right after

Two overlapping boot sequences each fetched config and called load(), which sets the list to "Loading…" then swaps in data — so on mount the board cycled Loading → data → Loading → data, the reported flicker/thrash. (The direct call also fired before the auth handshake.)

Fix

• Drop the redundant direct boot() on both PAGE (board) and NEW_ISSUE_PAGE — boot only via kit.initPluginView. The fallback kit still invokes the callback immediately when the DS kit can't load, so offline/degraded mode is unaffected.
• Harden load() with a monotonic token so a slow in-flight fetch that resolves after a newer load() started (rapid tab/filter clicks) drops its stale result instead of clobbering the fresh list — no thrash there either.
• Regression test test_pages_boot_once — each page boots via the kit exactly once.

87 tests pass, ruff clean. v0.1.3 (lockstep pyproject.toml + protoagent.plugin.yaml).

🤖 Generated with Claude Code

[protoquinn]

👀 Quinn is reviewing — verdict (PASS / WARN / FAIL) + findings to follow.

[protoquinn]

QA Audit — PR #14 | fix(board): boot once — stop the issue-list flicker on mount (v0.1.3)

VERDICT: WARN (CI non-terminal; leaning PASS on code)

---

CI Status

• test: in_progress

Diff Review

• Drops the redundant direct boot() call after kit.initPluginView(boot) on both PAGE and NEW_ISSUE_PAGE — the root cause of the double-config+load thrash (#13)
• Adds a monotonic loadSeq token to load() so a stale in-flight fetch that resolves after a newer load() started drops its result silently — clean defensive pattern
• Guarded the catch block's list.innerHTML assignment behind the same token check
• Version bumps pyproject.toml and protoagent.plugin.yaml in lockstep (0.1.2 → 0.1.3)
• Regression test test_pages_boot_once asserts each page has kit.initPluginView(boot) and no trailing boot();

Observations

• LOW: CI (test) is still in progress at review time. PR claims 87 tests pass + ruff clean locally; expect terminal green on re-check.
• The monotonic token pattern in load() is correct: pre-increment, capture, compare after await. The catch guard (if(my===loadSeq)) correctly prevents a superseded request's error state from overwriting the list. No gap.
• No unresolved CodeRabbit threads. No security-sensitive paths touched.

— Quinn, QA Engineer

[protoquinn]

Submitted COMMENT review on #14 — non-blocking while test CI runs; code is clean and the fix is correct. I'll re-evaluate for a formal PASS/FAIL when CI goes terminal.