With the piles sorted, spend the hours where they change verdicts. Refresh the reset intent against the three ECOs first, with the block owners in one room: no intent, no verdicts. Re-run the analysis against the variant's reset order, since that is the sequence the original signoff never checked. Then disposition each item into exactly one of four outcomes.
Outcome | When | Needs |
|---|---|---|
Fix now | Defective release synchronisation on a sequenced reset. | An RTL edit plus re-analysis; admit the milestone slips for these rather than waiving them. |
Waive with evidence | Current intent, clean re-analysis, and reset-aware simulation of the item's sequences. | The evidence package attached to the waiver, reviewed, with expiry. |
Conditional waiver | Evidence exists for the main variant but not the odd one. | Scope the waiver to the checked variants and flag the unchecked one as a known open risk. |
Refuse | Designer statement only, or stale intent that cannot be refreshed in time. | A written refusal with the containment: hold the milestone item open, do not sign. |
Expect the count to be uncomfortable: perhaps a handful waived, several conditional, and the rest refused or deferred. That distribution is the point. A triage that waives all twenty in 36 hours did not triage; it surrendered. Say which items you would fix even at the cost of the milestone, because reset release defects are boot failures, and boot failures are respins. The review-gate language for this stance is Pre-Tapeout Gates.
Refuse until intent is refreshed and analysis re-run: the simulation exercised a design the ECO may have changed, so its evidence is stale regardless of how green it looks.