A reconvergence report names a meeting point and two histories. Triage means reconstructing the design story behind those two paths: what source moment they share, how far apart they arrive, and what the consumer does with the mixture. The excerpt below is fabricated and labelled illustrative; real reports differ in layout but carry the same content.
illustrative report excerpt, fabricated for triage practice (not tool output)
finding R-014 rule: reconvergence of synchronised signals
convergence: u_dma_ctrl.cmd_en (AND of start_d and mode_d)
path A: src.start -> u_sync_start (2 stages) -> start_d, arrival cycle +2
path B: src.mode -> u_sync_mode (3 stages) -> mode_d, arrival cycle +3
skew: up to 1 destination cycle between the arrivals
consumer: cmd_en launches a DMA descriptor fetchRead the consumer first. cmd_en launches a fetch: a stale-mode mixture corrupts the descriptor, so this finding is high severity regardless of probability.
Reconstruct the shared source moment. Both paths start from the same source launch (start with its mode); the question is which mixtures the one-cycle skew can show.
Enumerate the reachable mixtures. With arrivals a cycle apart, the consumer can see new-start with old-mode and, on release edges, old-start with new-mode. Check each against what the consumer tolerates.
Decide fix or justify. A mixture the consumer cannot tolerate means restructure (next lessons); a provably benign mixture needs the insensitivity argument written down, not a silent waiver.
Record the evidence with the finding: protocol excerpt, waveform, or model run, so the next reviewer does not repeat the triage.
The severity drops: a one-cycle miscount is usually benign and bounded, so an insensitivity argument (error bounded to one count per launch) may close the finding honestly. The steps stay the same; only the consumer analysis changes the verdict.