Problem

v1 release preflight was failing in scripts/test_perf_budgets.sh on the verify_chain command budget.

The perf harness measured command budgets against state-100.json after also running a 500-repo scan in the same temp directory. Because proof chains are keyed by state directory, the 100-repo verify checks were run against a chain that already included both workloads, inflating chain size and causing false failures.

Changes

• Updated scripts/test_perf_budgets.sh to isolate 100-repo and 500-repo scans into separate workspace roots.
• Kept command-budget checks (score, verify --chain, regress run) tied to the 100-repo workload chain only.
• Preserved existing scan-duration checks for both 100 and 500 repo workloads.

Validation

• scripts/test_perf_budgets.sh
• scripts/run_v1_acceptance.sh --mode=release
• make prepush-full