anyAccessibilityDescendant

Whether any strict descendant of this element publishes an accessibility frame satisfying predicate. This element itself is never offered to predicate.

Unlike deepestAccessibilityHitPath this does not gate the descent on containing a position: the caller is asking about the subtree as a whole, not about one tap.

Currently unused in production. It was written for autograph-compose, which used it to ask whether a registered rect described a descendant of the resolved clickable rather than the clickable itself (#153) — a question that had to hold for taps anywhere on the clickable, so a position-gated walk would have answered it correctly only for some of them. That caller went away with the rect itself: #158 removed the descendant search, and #179 removed the whole geometric comparison in favour of observing whether trackClick's handler ran. Kept, with its tests, because the traversal is correct and independently useful; whether it should stay is #182.

Threading. Main thread only, for the same reason deepestAccessibilityHitPath is.

Termination. Bounded exactly as deepestAccessibilityHitPath is, against the same host-supplied and possibly cyclic tree: a branch revisiting a node already on the path is abandoned, descent stops at MAX_ACCESSIBILITY_TREE_DEPTH, and the walk stops after MAX_ACCESSIBILITY_NODE_VISITS nodes. Exhausting any of them returns false — which for the caller it was written for degraded to keeping the veto, i.e. a dropped event rather than a wedged main thread. A future caller inherits that bias and should check it suits them, matching how the rest of this file resolves the same trade.