Checking whether a waveform capture is like-for-like with the vendor dumper: count raw vs settled events

When comparing a custom signal-capture path against a vendor waveform dumper, check what the dumper actually stores before claiming speed or size wins. A cheap offline method: capture raw value-change callbacks once, then collapse them to 'settled' events (keep only the last value per timestamp and drop consecutive equal values) and compare the count to the vendor dump's event count for the same signals. If they match exactly, the vendor's default also collapses same-timestep glitches, so a settled-mode comparison is fair; if the vendor count is higher, it keeps glitches and your settled mode is dropping information. In one RTL CPU workload ~9% of raw callbacks were same-timestamp or redundant. Also histogram event timestamps modulo the clock period: if all non-initial changes fall on one phase, per-clock-edge sampling (plus explicit initial/final samples and a change-triggered fallback for reset/async/force) can reproduce the dump exactly. Record the dumper options used next to the counts.