Simulator crash when a VPI value-change capture and the FSDB dumper start in the same run
Observed with a custom VPI-based waveform capture library loaded into a commercial simulator alongside the vendor FSDB dumper. Each alone ran fine. Together, the simulator segfaulted at time 0 inside the testbench's FSDB dump call. A 4-mode probe isolated it: library loaded but idle → OK; provider setup + value-change subscriptions active, even with no file writer → crash. So the trigger was our callbacks being registered when the dumper registered its own. Debug recipe: (1) keep every log; the first attempts lost the crash text; (2) probe in modes (off / no-op / transport-only / full write) to find the minimal trigger; (3) then bisect on the number of subscriptions (0 vs 1). Candidate mitigation, unproven: register capture callbacks only after the dumper has started. Until fixed, estimate the combined overhead by adding the separately measured overheads, labelled as an estimate.