click and press report dispatch, not landing: state the contract and track the outcomeObservation gap #3335

Purpose

A successful click or press means the tap was dispatched. It does not mean the tap landed on the intended element. Neither the help text nor the guarantee matrix states this, so callers and lane tooling read click-ok as action-landed.

This issue states that contract where callers read it and gives the gap a live tracking issue. It does not add a landing check to the default click.

Evidence

On main run 37815426076 (2026-10-08, iOS 26 lane), 01-settings.ad attempt 1 tapped (201,319) and reported success. The settled General cell sits at (201,406), which is where attempt 2 tapped. The next wait failed with zero readable captures, so the retained evidence cannot say what the tap hit. Full timeline: #2948.

The replay behaved correctly here. Step 4's wait is the landing check, and it failed instead of letting the replay continue.

Decision

Click and press report dispatch, not landing. If a caller needs landing evidence, it uses --verify or --settle, or follows the action with wait/is.

Rejected alternatives:

• A bounded post-action probe on every default click. It would add a capture to every click: a warm iOS click is about 2.6 s, and 2.0 s of that is already the snapshot, so click time would roughly double. It would also fail when it matters most. The failing wait got no readable capture in 5 s, so a probe with a smaller budget would come back empty on the same runs. Finally, it duplicates the existing --verify and --settle options.
• An activation: "unobserved" field on every response. On the default path the field would always have the same value, so it gives callers nothing. Add a field only when some path can report a different value.

Scope

The gap is not iOS-specific. The cell that owns it is outcomeObservation, not verifyEvidence. verifyEvidence is the opt-in --verify feature, and it already works. TAP_OUTCOME_NOT_OBSERVED_GAP (packages/contracts/src/interaction-guarantees.ts) already waives outcomeObservation on the runtime-selector and runtime-ref paths. DIRECT_IOS_OUTCOME_NOT_OBSERVED_GAP does the same on the direct-iOS paths. Both still point at the closed umbrella #1081.

Out of scope, tracked separately:

• The stale tap point. On this run the tap point came from a capture taken about 0.5 s after open, while the list was still settling (postOpenObservation: unobservable). That is a separate pre-action defect: iOS selector click taps a point resolved from a pre-tap capture, so a layout shift between capture and tap retargets it silently #3354.
• The wait observation stall: ci(ios): the iOS simulator smoke lane is failing on main across several rotating signatures #2491/ci(ios): a regular depth-1 snapshot carries XCTest tree quality metadata when the AX bridge probe circuit is open #3328.

Required behavior

• click and press help, plus the user docs, state that success means dispatched, not landed, and name --verify, --settle, and wait as the ways to confirm landing.
• TAP_OUTCOME_NOT_OBSERVED_GAP and DIRECT_IOS_OUTCOME_NOT_OBSERVED_GAP point their trackingIssue at this issue instead of Interaction guarantee gaps (ADR 0011 umbrella) #1081.
• The response shape, the default click budget, and the replay retry policy stay as they are.

Completion

• Versioned help for click/press and website/docs/docs/commands.md state the dispatch-only contract.
• Both waivers reference this issue.
• The issue stays open as the tracking issue for the waived outcomeObservation cell until a path can observe the outcome on its default success path.

Related: #2948 (diagnostic run producing this evidence), #2491/#3328 (wait observation stall), #2343 (wait evidence fields), #1081 (closed ADR 0011 gaps umbrella).