^
OK …
Though we do know that users say they play ahead of the click in order to get the sound to emerge together with the click.
How about this? …
Suppose someone did another test, where - using the same PC, etc - they recorded in time with the click, but while listening to a sound unaffected by latency - eg using a keyboard instrument that outputs its own sounds.
And, further, suppose it were found - consistently, over several recordings - that (a) when recording using a zero-latency instrumental sound, the MIDI was recorded properly on the beat, but (b) when recording using an instrument with latency, the MIDI was logged too early.
It seems to me that that would provide quite strong evidence that the reason for the MIDI being recorded early was that the performer was anticipating the beat in order to make the sound come in time with the beat.
(While doing the recording using a zero-latency sound, the VSTi ought to be still operational in Cubase, so as to eliminate its presence/absence as part of the cause of any observed effect.)