Some grammars, e.g. Opperman 1967, give azi as a future marker as well (i.e. epistemic certainty). I used to think this is a misanalysis, arising from confusing azi 'to know' with future particle zo (often reduced to clitic z). But given that I found an authentic sentence with negative future with azi (see Example 54), maybe there's something in the idea of future deriving from Zulu azi 'to know, be able' as much as from Zulu auxiliary future marker zo.
For epistemic possibility the "perhaps" reading is not possible; but a contextual reading of the other sort proposed is just about possible.