Ok, here are new findings :
When the dropouts are occurring, tweaking the core affinity for Cubase.exe in Task Manager (for example, disabling one core, doesn’t matter which one) makes the dropouts go away completely !
However, as soon as we tweak the buffer size, the dropouts appear again, which is again fixed by tweaking the core affinity (even if we check back all cores).
Tweaking the core affinity seems to “rearrange” the instructions between the cores so they do not clash any longer, but as soon as we touch the buffer (which resets the audio engine) it becomes buggy again.
Please tell me how strange is this !
Also, if we toggle Multi Processing back and forth during the “affinity fix” (when at least one core is disabled), the issue comes back but this time it’s glitching out even more than before, sometimes it makes long glitching noises and after a while it completely drops out for a brief moment (just like when changing the buffer / audio engine is reset), and when it does that the current MIDI note will hang, like if Cubase (or Spitfire ?) was trying to use the disabled core.
This is 100% process scheduling fault right here.
It now remains to be determined whether it comes from Cubase, Spitfire, or Windows.
I have sent a ticket to Spitfire and put a link to this topic.