flaky perry-stdlib test: thread_exit_releases_the_threads_fresh_symbol_pointers ('a dead thread's Symbol() stayed in SYMBOL_POINTERS', ~1/40 on Linux) #11539

runtime_thread_exit_tests::symbols_tests::thread_exit_releases_the_threads_fresh_symbol_pointers failed once in 40 runs of the full perry-stdlib lib test binary on Linux (perrymaster, CI env: CARGO_PROFILE_TEST_DEBUG=0 CARGO_PROFILE_DEV_DEBUG=0 CARGO_INCREMENTAL=0, default test threads). It exited 101 with:

a dead thread's Symbol() stayed in SYMBOL_POINTERS

It showed up while validating #11538, which touches only the fetch lifecycle tests. The other 39 runs passed.

Unverified hypothesis: a concurrently running test's Symbol() reuses the freed address, so the membership assertion sees a live entry at the same address rather than the dead thread's pointer. If so, the assertion needs an identity check (e.g. a generation or owner tag), not just an address test. Otherwise it is a real release gap in the thread-exit hook.

Impact: it can intermittently redden #11424, which adds the perry-stdlib lib tests to the per-PR runtime-change fan-out.