we should say that it stars on 3s and ends on 1s
This is correct. We dont use per second intervals, it would make discussing frame modulation very confusing.

Incidentally, all these scenarios would be related to timing, not spacing. Spacing, at least to how I feel is best, describes the drawings themselves. For instance you can animate on 2's and keep the drawings very close together, or you can animate on 2's with more exaggerated poses with large gaps in the arc of motion.