Map each applicable requirement to the evidence needed to check it—such as tests, assertions, reference-model comparisons and coverage—and record exclusions or waivers separately. A coverage bin alone shows observation, not correctness; a waiver records an accepted gap rather than verifying the requirement.
No. First establish the maximum admitted work and the observation capacity required to retain it. An overflow invalidates the checker's evidence for that interval; it is separate from a protocol deadline or a proven DUT fault. Increase capacity or constrain the exercised subset explicitly, then rerun with nonzero observed work.
False fail rate: scoreboard key too strict or model too ideal.
Escape: scoreboard key too loose or missing scenario.
Regression: preserve each fixed bug with the smallest reliable reproducer and suitable checker, which may be an assertion, directed test, model comparison or another automated check.