# clutter_endurance_500_restore_reemission

Result:

- The replay succeeds `500/500`.
- `compile_100_a` and `compile_100_b` are identical in both `bundle_text` and `included_items`.
- `compile_250_a` and `compile_250_b` are identical in both `bundle_text` and `included_items`.
- `compile_500_a` and `compile_500_b` are identical in both `bundle_text` and `included_items`.
- `compile_500_a` matches `compile_250_a` in both `bundle_text` and `included_items`.
- `compile_mutated` differs from the checkpointed compile and contains `src/gateway/proxy.rs`, `leave temporary bypass enabled`, and `allow unsigned replay until Monday`.
- The final restored bundle contains `OMEGA-9921`, `src/auth/webhook_guard.rs`, `re-enable HMAC verification`, and `reject unsigned webhook replay`.
- The final restored bundle excludes `src/gateway/proxy.rs`, `leave temporary bypass enabled`, and `allow unsigned replay until Monday`.
- The bounded token estimate is `449` at step 100, `566` at step 250, `580` for the mutated compile, and `566` at the final restored compile.

Interpretation:

This case shows a bounded endurance extension rather than a new benchmark family. Two retained target items are seeded at the start of the run, valid clutter continues for 500 replay steps, compile remains deterministic at the checked stages, checkpoint and restore return the runtime to the checkpointed relevant state, and the final bounded compile still re-emits the retained auth-fix subset instead of the long clutter history.