Slice 11b.1 W1 retest hypothesis (P0 #4): AstroHan reported "向上滚动之后
就会不断向下滑" — scrolling the timeline upward immediately gets snapped
back to bottom. Pure-reactivity RCA across `createAutoScroll`,
`use-session-scroll-dock`, the timeline scroll controller, and the W1
leaves did not surface a deterministic code path that would override
`userScrolled`. The remaining suspect is a same-frame race: handleWheel
synchronously calls `stop()` (which writes `userScrolled = true` via
`setStore`), but the streaming-time `createResizeObserver` callback can
fire in the same frame *before* Solid propagates the store read inside
this hook, so `scrollToBottom(false)` lands while the gate still thinks
the user is at bottom.

Belt-and-braces fix: stamp `lastUpwardGestureAt` inside handleWheel and
short-circuit the resize-observer auto-scroll for 500ms after any
upward gesture. The window covers a single wheel tick + layout pass
while staying short enough that users can drop back into auto-follow by
scrolling to the bottom themselves.

No behavioural change while the user is idle or scrolling downward; the
existing isAuto / userScrolled / canScroll gates still apply. The fix is
intentionally local to the resize observer — handleScroll keeps its
threshold-based path so the auto-snap on real bottom-pin restores still
work as before.

…d frame

AstroHan W1 second retest flagged 3 structural regressions on the W1
leaves; this commit fixes the trio in one atomic change because all
three are CSS-only.

1. **Text selection regression (#1)** — The app shell sets `select-none`
   on the timeline body (layout.tsx L2320) and opts back in for input,
   textarea, and `[contenteditable]` only. The legacy carve-out at
   message-part.css L710 named `[data-component="user-message"]
   [data-slot="user-message-text"]` and `[data-component="text-part"]`
   so legacy bubble + assistant prose stayed selectable; the W1 leaves
   replaced both data-components without an equivalent carve-out, so
   bubble text and agent prose stopped responding to drag-select.
   Reinstated by adding `user-select: text` + `cursor: text` to
   `[data-slot="bubble-text"]`, `[data-slot="agent-prose"]`, and
   `[data-slot="agent-reasoning"]`.

2. **Trow chev direction (#3)** — The W1 icon is `chevron-down` (rest
   reads "click to expand"), but the open-state rotation was 90°,
   which pointed the glyph sideways. Per AstroHan's flag the open
   chev should read "click to fold up" — 180° rotates `chevron-down`
   to chevron-up, restoring the disclosure idiom.

3. **Trow expanded nested frame (#4)** — DESIGN.md L417 scopes the
   transparent + 1px `--border-weaker` + radius-sm body frame to each
   per-tool result body, not the trow group wrapper. The W1 trow-block
   wrapped the inner sub-rows in its own frame, which composed with
   the legacy per-tool chrome and produced a visible double border.
   Dropped the wrapper's border + padding so it only owns the vertical
   rhythm between sub-rows; the per-tool body keeps its own frame.

Tests still pass: `session-turn-user-bubble`, `session-turn-agent-round`,
and `session-turn-trow-block.reducer` suites are CSS-grep + reducer
based and do not assert any of the changed declarations.

AstroHan's third W1 retest (msg=362e9b72) raised two visual issues
on the trow block; both land in `session-turn-trow-block.css`.

**#2 — per-tool box wrapper.** Slice 11b.1 #5 B+ added a transparent
+ 1px --border-weaker frame around each tool's rich body
(`data-slot="trow-result-body"`) as the L417-spec scope marker.
AstroHan reported the chrome makes tools look like isolated cards
and (when bash is expanded) stacks on top of bash-output's own
border. The W1 intent is a flat stacked feed where tools 叠在一起.
The wrapper stays — it is still the scope boundary for the inner
chrome typography reset — but its visual chrome (border / radius /
padding) is removed.

**#4 — chev direction + inner tool chev size.** Two parts:

  (a) Trow summary chev direction. The third retest re-defined the
      affordance: 折叠朝右、展开朝下. The second-retest fix that
      pointed the chev up on `details[open]` is reverted —
      file-tree, Collapsible, and the W1 mocks all use down-when-
      open, so trow should match. Default `transform: rotate(-90deg)`
      (chevron-down → right when collapsed); `details[open]` flips to
      `rotate(0)` (natural down).

  (b) Inner Collapsible.Arrow chev size inside the trow result body.
      AstroHan flagged that the chev inside each tool trigger ("执行
      命令 XXXX" row) renders as a "大 chevron" — Collapsible's default
      is 16px, but the W1 caption surface (DESIGN.md L412 lock) caps
      the chev at 12px. Scoped reset fires only inside
      `[data-slot="trow-result-body"]`, so the Collapsible default
      elsewhere (file-tree, settings, etc.) is unaffected.

…te thinking heading

Two related AstroHan W1 retest 4 issues (msg=ac13481a) on the
"思考中" surface and the reasoning-summaries setting.

**#4 — Thinking heading overflows right edge.** When
`showReasoningSummaries` is off the thinking row renders a
`<TextReveal>` showing the latest reasoning heading. `TextReveal`
defaults to `white-space: nowrap` on its entering/leaving spans, so
a long heading bleeds past the column's right edge. The component
already exposes a `truncate` prop that caps width and adds
`text-overflow: ellipsis`; opt in.

**#5a — Reasoning-summaries setting had no effect on COT body.**
The settings toggle ("显示推理摘要 / Show reasoning summaries") flows
through to `SessionTurn` but only gated `showThinking`'s assistant-
visible counting and the secondary reasoning heading inside the
thinking row. `AgentRound` never received the prop, so reasoning
groups always rendered regardless of the toggle. Pass
`showReasoningSummaries` to `AgentRound` and drop `reasoning` groups
from the rendered output when it is explicitly `false`. The grouper
itself stays structural (does not know about user preferences); the
filter lives at the leaf consuming the groups, matching the existing
seam with `partState()` in `session-turn-helpers.ts`.

…hine

State machine: pending → resolved (tombstone, 30s TTL) → cleaned.
Exports: register, lookup, resolveIfPending, failIfPending, hasPending,
onSessionDestroyed. Cleanup is on-lookup (no timer thread); the clock
is injectable for tests.

Semantic split (addresses v10 round-3 P2 #4): onSessionDestroyed fires
only on server shutdown / session delete and rejects pending Deferreds
with `ExternalResultError({reason: "shutdown"})`. User turn aborts will
flow through ctx.abort → failIfPending with reason: "aborted" — wired
in a later commit. The two paths never overlap; reason is always
"aborted" XOR "shutdown" XOR completed-normally.

Registry semantics (addresses v10 round-2 P2 #1): hasPending counts
only pending entries (tombstones excluded); resolveIfPending returns a
discriminated outcome ("resolved" / "already_resolved" / "not_found")
that the route maps to 200 / 409 / 404. Within the 30s tombstone window
a retry of the same user gesture sees the deterministic 409 instead of
a 404 race.

No consumers yet (route, ctx.externalResult, session-destroy hook all
come in later commits).

Part of #756 PR A.

Wires ExternalResult.onSessionDestroyed into the existing
Session.clearPendingInteractions path that already tears down the
Question/Permission/SessionBlocker services on session delete and
session archive. Pending external-result Deferreds get rejected
with ExternalResultError({reason: "shutdown"}); tombstones are
dropped.

Per v10 semantic split (round-3 P2 #4): this hook does NOT fire on
turn abort or status transitions. Turn aborts route through
ctx.abort → failIfPending with reason: "aborted". So `reason:
"aborted"` and `reason: "shutdown"` never overlap.

Part of #756 PR A.