For vBET cache tables we choosed utf8_bin comparator, because we believe this is faster than utf8_general_ci. It is simpler - no rule just compare bytes, so it should be faster that any other comparation based on additional rules (not tested - just based on out logic).
Also we cannot use use utf8_general_ci, because it is canse insensitive, and we need exact matches in cache.

Please note that we do not change anything in vB model, so I do not exactly understand why you are asking about not converting db table charset... We do not convert anything in DB. vBET creates its own tables for cache and do not touch existing ones - nothing is converted there...