I’ve only seen diviners as a type handled properly in one game which, if my faulty memory is correct, was C&S (2nd edition, probably also 1st, but not 3rd onward), but I may be huffing paint thinner.

First, diviners had a group of useful spells like detecting traps, hidden things, etc. This meant you didn’t get that whole weirdo vision quest thing with “information” that was only recognizable as such long after the fact, rendering the divination kind of useless, as the only thing a diviner could do.

But even for the visions there was a decent system in place. The diviner would cast the spell and based on the results of that roll, paired with a roll (or decision) made in secret by the GM, get a degree of success that translated into percentile points. The GM’s roll/decision decided between good or bad omens.

If the GM rolled/decided on good omens, they’d come up with one of those vague, flavourful visions so beloved byirritating to players. But… at any point for the duration of the cast augury, if something that could kinda/sorta be interpreted to belong to that vision showed up in play (GMs being encouraged to err on the side of the player), the player could use some of that percentile pool to modify die rolls in their favour (or, equivalently against the opposition’s favour) to do things like turn failures into successes, or successes into critical successes or the like, thus retroactively making the vision “come true” mechanically.

If the rolled/decided omens were bad omens, the percentile pool (smaller if the player rolled well, larger if the player rolled poorly) was instead given to the GM to use to stymie and confound the players in ways related to the vision.

The end result was that the flavourful vision was there, but its application to the situation was determined in play and had mechanical relevance, which was satisfying to the players.