Private brand check #x in o returns false for real instances #9411

class A { #x = 1; static has(o) { return #x in o } }
A.has(new A())   // node: true   perry: false

Also false for #method brands and for subclass instances.

Per ES §13.10.2, #x in o is true when o has the private field installed — which new A() does. Perry answers false, so downleveled TypeScript/Babel brand checks silently take the wrong branch: code that asks "is this one of mine?" concludes no, and falls through to whatever it does for foreign objects.

Silent wrong answer, no error — the failure mode is a wrong code path rather than a crash.

Found by a differential stress-test of claude-code under perry.