I must admit I just triggered some samples, so nowhere as much exactness needed as in your use case, but I think I too encountered some funkiness with missing triggers.
since then I didn’t manage to check back

your documentation in that lines thread is great, thanks!
btw: dev=1 changing something might indicate that the midi logic is using too much (teensy-) resources, since the dev flag just limits everything to fader1 for debugging purposes (well I saw some mixture with the channel count and midi channels in at least one of the midi branches).
I guess you don’t have the michigan with switchable pullups… I do and needed to limit some functionality to even fit ram (its teensy LC is reduced and the midi fork adds a lot of realtime processing). so that could indicate some lack of headroom in my setup, but IIRC I got that down to comparable figures first by limiting midi channels.
I need to get into this again and run more tests also with midi notes…