[framework-context] Container registration benchmarks exceed repository verification thresholds on unchanged source #3435

Problem

The local repository verification gate fails its Container registration/validation benchmark thresholds while resolving #3006. The failing benchmark and Container implementation are unchanged by that fix.

Evidence

Base: 02b2e262ea083076b613839cf2749ee3c44c1495.

• pnpm check: Container.register × 50 components p75 89.5ms > 10ms; Container.validate (50 components) p75 26.9ms > 20ms.
• Separate pnpm bench:check on the fixed candidate: registration p75 42.4ms > 10ms.
• Replaced only Context.ts with the exact base production file, ran the same benchmark command, then restored the candidate: registration p75 60.5ms > 10ms; validation p75 30.8ms > 20ms. All other production code, benchmark files, dependency metadata and thresholds are byte-identical between the candidate and base. Regression spec and changeset are not benchmark inputs.

Other repository sessions were running heavy checks on this host. The evidence establishes that the failure reproduces without #3006; it does not establish whether the cause is contention, a benchmark harness problem, or baseline performance debt.

Desired outcome

Investigate the registration/validation benchmark gate on the unchanged source, distinguish host contention from a repeatable performance regression, and obtain trustworthy passing verification without hiding failures or relaxing thresholds merely to pass.

Scope

Keep this separate from request cleanup Problem serialization in #3006. No benchmark implementation or threshold was changed in that fix.