Parent: #498 Related: #1074, #1076, #1077, #1078, #491
The live inventory labels the polymorphic property loop as legacy_pattern_kernel / legacy_narrow. It recognizes one exact 38-bytecode indexed-selector/update/checksum body, then batches selected-object property reads/writes, numeric arithmetic, and loop control. A compiler candidate hint admits the body. This conflicts with the no body/benchmark recognizer constraint.
Executed baseline on a6a984e7: the kernel-hit witness passes 2/2. In both synchronization modes, ordinary objects return 2,498 with 1,287 steps, and the accessor variant returns 1,602,498 with 1,485 steps. Both results match Node. The witness asserts hits in isolated execution and no kernel hits in synchronized execution.
Acceptance:
- Remove the polymorphic-loop candidate hint, exact body recognizer/executor, plan type/tag, and hit metric; retain general polymorphic property inline caches, ordinary indexed operations, and shared helpers still needed by live families.
- Preserve both baseline results/step counts in both modes and ordered property/receiver effects, shape/descriptor/prototype changes, coercion, and partial effects before exceptions.
- Cover actual no-GIL workers, explicit moving-GC roots, source hooks, budget/stop checkpoints, and allocation failures.
- Focused normal/ReleaseSafe/unsuppressed TSan checks and exact before/after affected test262 and tree/required-VM comparisons show no new failures.
- Regenerate live inventory/docs/manifest; preserve historical artifacts and explicit labels for remaining legacy work.
This deletes body-specific work substitution; it is not current throughput evidence or completion of #498.