For completeness, there was an almost similar discussion here:
Proposed solution : rendering the external gear audio before using vst connect. This will work without a doubt, as soon as tracks stabilized enough.
While I understand that VST Connect is not a jamming tool, the point is that FMPOV midi tracks to external gear and the audio returns with the loudspeaker symbol switched on could be prebuffered by Cubase just as it is done for VSTi instruments, couldn‘t it?
I have been successful in the meantime by delaying all VSTi tracks by e.g. 500 msecs (if this is the configured remote delay time of vst connect) to have everything in perfect sync, but I have to switch it off again when not being connected anymore.
And this workaround has additional drawbacks, e.g. when cycling the first 500 msecs the delayed tracks are not playing anymore, same applies for the end of the cycle.