release: pre-existing Linux parity failures block v0.5.1519

[proggeramlug]

Release evidence

The mandatory Full CI for candidate 32cc6ea surfaced three Linux parity failures that are not currently tracked in test-parity/known_failures.json:

• test_stress_promises: Perry SIGSEGV (exit 139).
• test_class_field_layout: output mismatch; both runtimes exit 0.
• test_issue_3199_3200_tls_server_tlssocket: Perry exits 1 after the initial server-address line; Node exits 0.

Full candidate run: https://github.com/PerryTS/perry/actions/runs/32898410858

Focused five-attempt controls show these predate the #8835 batch:

• Promise stress on pre-batch: land #8825, #8826, #8830, #8832 #8835 r13 (88583f0): 5/5 SIGSEGV: https://github.com/PerryTS/perry/actions/runs/32911025459
• Promise stress on r14 (32cc6ea): 5/5 SIGSEGV: https://github.com/PerryTS/perry/actions/runs/32911028305
• Class-field layout on r13: 5/5 output mismatch: https://github.com/PerryTS/perry/actions/runs/32911407743
• TLS server/TLSSocket on r13: 5/5 Perry exit 1: https://github.com/PerryTS/perry/actions/runs/32911412179
• TLS server/TLSSocket on r14: 5/5 Perry exit 1: https://github.com/PerryTS/perry/actions/runs/32911414409

test_parcel_watcher_idle was also surfaced, but it is not a standalone parity program: it requires PERRY_WATCH_ROOT, PERRY_WATCH_READY, an external watcher package for Node, and an orchestrating harness. The release unblocker classifies it with a parity-skip header instead of recording a product bug here.

Acceptance criteria

• test_stress_promises runs to completion and matches Node under repeated Linux release builds.
• test_class_field_layout matches Node byte-for-byte.
• test_issue_3199_3200_tls_server_tlssocket completes with the same lifecycle/output as Node.
• Remove each corresponding known_failures.json entry in the same PR that fixes it.

[proggeramlug]

Release candidate r16 surfaced one additional Linux parity mismatch before shard 8 hit its 150-minute timeout:

• test_parcel_watcher_facade: Node exits in package_json_reader; Perry receives native events but times out waiting for the nested create event.

Evidence: https://github.com/PerryTS/perry/actions/runs/32922022811 (parity shard 8, 152/173 tests before timeout).

Tracked as a Linux ci-env known failure in #8846, which also moves parity from 8 to 12 shards and raises the timeout backstop so release-grade Full CI can complete.