[Bug] TodoDock fires duplicate session-todos read on mount

[muzimu217]

[Bug] TodoDock fires duplicate session-todos read on mount (0.15.10, main)

Problem / 问题

TodoDock triggers the session-todos read twice on mount for the same session, instead of once:

1. useSessionTodosRecovery(sessionId) → startSessionTodosRecovery() → initial refresh() (recovery read);
2. TodoDock's own mount effect → todosGet read (feature read).

Both hit todosGet for the identical sessionId on every dock mount/session switch. The recovery path was introduced with the checklist-recovery feature, so the duplicate is new behavior on current main.

Why it matters / 影响

• Every session switch/activation costs a duplicate IPC round-trip to the host;
• The recovery read and the feature read can race — the store revision fence protects correctness, but the duplicate still wastes host work and can produce flaky ordering in tests (a freshly-added test currently expects exactly one read on mount).

Suggested fix / 修复建议

Either:

• Let useSessionTodosRecovery own the mount read and have TodoDock skip its own initial fetch when recovery is active; or
• Deduplicate inside the store: applyTodosChanged/read wrapper drops the second in-flight todosGet for the same sessionId+revision window.

Reproduction / 复现

node --test apps/desktop/test/todo-recovery.test.mjs — the test TodoDock recovers a failed first read when the host returns without a session switch currently asserts reads deep-equals ["session-a"] after the first render+settle and fails with 2 !== 1 on current main (verified on a pristine checkout of e77404ea3, no local patches).

Note: this failure also means current main carries a red test in the desktop suite — flagging since the contributing CI workflow may not have run for the original PR.

[yexisu]

已在当前 main（3e34ca07d）验证，该问题无法复现，测试全部通过：

node --test apps/desktop/test/todo-recovery.test.mjs
ok 1 - TodoDock recovers a failed first read when the host returns without a session switch
ok 2 - TodoDock re-reads cached sessions on activation and restart without snapshot-driven requests
ok 3 - TodoDock keeps newer events and drops old-host and switched-session reads
ok 4 - TodoDock never reads the local host for remote or native sessions
ok 5 - TodoDock rejects a mismatched snapshot and ignores pending work after unmount

报告中断言 reads 深度等于 ["session-a"] 的用例（首个 render+settle 后恰好一次读取）现在通过，说明挂载时只发起一次 todosGet 读取，双重读取已不存在。TodoDock 自身也没有独立的挂载读取（挂载读取由 useSessionTodosRecovery 统一负责）。先关闭；如在新版本上仍能观察到重复 IPC，欢迎带版本号重新打开。

[xpeng5278-web]

I’d like to take this issue. I’ll trace the TodoDock mount and recovery read paths, then make the smallest change that ensures one session-todos read per mount while preserving recovery behavior. I’ll run the focused todo recovery test before opening a PR.

[muzimu217]

补充一个独立验证结果，供关闭参考：

• 上游已在 main 修复该问题（c744f7c39 "recover authoritative checklists after host restart" 引入 recovery 去重，3e34ca07d 已含）。
• 我在本机独立复核：node --test apps/desktop/test/todo-recovery.test.mjs 全绿（5/5），真机 E2E TODO_CHECKLIST 6 项检查含 initial-read-failure-ready-recovery 与 sqlite-host-restart-cached-snapshot-recovery 均 ok。
• @xpeng5278-web 提醒：当前 main 已无法复现 mount 双读，接单前建议先在最新 main 复跑同一测试确认，避免在已修复面上做改动。

若维护者确认，本 issue 可按 completed 关闭。