nth dropped when storing aria snapshot nodes

The nth field from each AriaSnapshotNode is not included in the mapping, so it is never stored in state.ariaSnapshotNodes. When refLocator resolves an ax<number> ref whose original node had nth set (used to disambiguate multiple elements sharing the same role+name), it falls through to the plain getByRole(role, { name, exact: true }) call without calling .nth(...), silently returning the wrong (first) element.

This is a comment left during a code review.
Path: extensions/browser/src/browser/pw-tools-core.snapshot.ts
Line: 65

Comment:
**`nth` dropped when storing aria snapshot nodes**

The `nth` field from each `AriaSnapshotNode` is not included in the mapping, so it is never stored in `state.ariaSnapshotNodes`. When `refLocator` resolves an `ax<number>` ref whose original node had `nth` set (used to disambiguate multiple elements sharing the same role+name), it falls through to the plain `getByRole(role, { name, exact: true })` call without calling `.nth(...)`, silently returning the wrong (first) element.

```suggestion
    nodes: formatted.map((n) => ({ ref: n.ref, role: n.role, name: n.name || undefined, nth: n.nth })),
```

How can I resolve this? If you propose a fix, please make it concise.

Misleading test name — says "does not flag" but asserts the opposite

The test is named "does not flag inconsistent_timestamps when startedAt is captured before task registration", yet the assertion expects the inconsistent_timestamps finding to be present. The comment clarifies the intent (testing legacy pre-clamp data), but the name contradicts the assertion and will confuse future readers. Consider renaming to something like "flags inconsistent_timestamps for legacy data with startedAt earlier than createdAt (pre-clamp)".

This is a comment left during a code review.
Path: src/tasks/task-registry.audit.test.ts
Line: 129-153

Comment:
**Misleading test name — says "does not flag" but asserts the opposite**

The test is named `"does not flag inconsistent_timestamps when startedAt is captured before task registration"`, yet the assertion expects the `inconsistent_timestamps` finding to be present. The comment clarifies the intent (testing legacy pre-clamp data), but the name contradicts the assertion and will confuse future readers. Consider renaming to something like `"flags inconsistent_timestamps for legacy data with startedAt earlier than createdAt (pre-clamp)"`.

How can I resolve this? If you propose a fix, please make it concise.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

Ignore stale frame selector for ax refs

This new ax<number> branch scopes lookups through state.roleRefsFrameSelector, but storeAriaSnapshotNodes never clears that field. If a user takes a frame-scoped role snapshot first (which sets roleRefsFrameSelector) and then takes an aria snapshot, later refLocator("ax...") calls are forced into the old frame and fail to find elements that are in the top document. That makes valid aria refs intermittently unusable depending on prior tool calls.

Useful? React with 👍 / 👎.

Preserve disambiguation when caching ax snapshot refs

The cached aria node metadata now keeps only {ref, role, name}. For pages with repeated accessible names (for example multiple button elements named "Submit"), different ax refs collapse to the same getByRole query, so click/fill paths can hit Playwright strict-mode ambiguity instead of the exact node the snapshot ref pointed to. refLocator already supports node.nth, but this mapper never populates it, so the disambiguation path is unreachable for real aria snapshots.

Useful? React with 👍 / 👎.