In that post I was talking about the selections, not the notes within the selection. More specifically, it seems selecting by time (measures) rather than notes is necessary to carry out groove quantize. I'm theorizing groove quantize to have some analogy to paste, which is a little easier to talk about. Selection and snap value are important for copy-paste.
If you make a selection by notes, say in PRV, and copy-paste them, the first selected note will paste to the snap offset of your now time. If your first note was originally deliberately grooved off-beat (off-snap), copy-paste would change the timing of your first note to land on a snap, not on your intended groove. All the rest of the notes would be off too by the same amount. The only way to paste that selection of notes with the right timing would be to turn off snap and put your now time at the
tick offset matching the first note of your selection.
So that would suggest selecting by time is required if time alignment is important. Select by time goes by snap boundaries. Then the copied notes will remain "grooved" within the time selection and then will land with their intended offsets to your snap when they are pasted.
One way to observe this with copy and paste is to record a few notes off-beat. This creates a clip delimited by the first and last notes. Then copy this clip and paste it. The clip lands on-snap, and let's say snap is at a beat, so the first note of the copy is now on-beat instead of the original off-beat performance. For the alternative, select a snapped time interval containing the original notes. Or slip-edit the original clip to snap boundaries. Then copy and paste. This time the off-beat timing is preserved.
Considering "groove quantize" now, as opposed to paste, the "groove" is applied starting at the beginning of the target selection. So it would seem best that both the target selection and the source selection for the original copy start at measure boundaries. What's more, if the copied selection does not represent a whole number of measures, the whole idea of copying a groove over multiple measures wouldn't make sense. I said "measure boundaries" in my previous post because grooved performances are typically meant to be relative to measure boundaries.
Theoretically, some kind of groove quantize might still work if both the target selection and source selection started at the same relative
tick offset within a measure. But that's hard to think about and hard to test.
I hope this clears up any misunderstanding from my previous post. I couldn't tell from the video if your selections were by time. I think you selected a clip, but I didn't know the time alignment of that clip. So I don't really know if that would make a difference in what you wanted to get. In the past I have made grooves from audio clips and applied them to MIDI, seemingly with success.
Bill B.