Those problem are probably related to this (search for “delay”):
https://lore.kernel.org/lkml/be30446c-f350-471d-bfac-b4b8dc0a75a2@starfivetech.com/
(The M.2 SSD 970 EVO Plus mentioned in the OpenBSD mailing list is a 250GB/500GB/1TB or 2TB SSD with 2GB of LPDDR4 SDRAM which supports PCIe 3.0 x4 - Idle (ASPT on) 30 mW ; Active (Avg.) Read 5.5W ; Active (Avg.) Write 6.0W - so under average usage it could potentially require 11.5 watts or at 3.3 volts nearly 3.5 amps of current, I can not find any information about peak power requirements, but that could be very high indeed for a few microseconds to milliseconds)
At a guess the very large and very fast (very higher power) NVMe devices draw so much current that they are glitching the power and clocks of the VF2, and it needs an extra delay beyond what the specification suggests for them to both stabilise before the NVMe can be accessed.