I’ve always liked the Ars Magica stress dice mechanic.
Roll a d10 – a zero indicates a potential problem, a one doubles the next roll (where a subsequent 0 is a 10) and anything else is the value given.
If a zero is rolled, you then get to roll a number of botch dice according to how risky the action was, a single die if it was not particularly risky, up to dozens or more (botching a casting roll in a divine aura when using raw vis to boost your spell casting *8′). If a botch die comes up zero, something has gone badly wrong – the character has actually botched. If you rolled more than one botch dice and more than one of them came up zero multiple zero’s you have the potential for a double botch or triple botch etc. which allow for more extreme failures. If non of the botch dice come up zero, then you’ve been lucky and you just got a zero on the roll – which may still be enough to succeed if your stat/skill is high enough.
It doesn’t have the ugly discontinuities that the shadow-run exploding dice mechanic has (what use is difficulty 6?), it allows botches to be 1 in a 100 or worse depending on risk and very occasionally allows really quite awesome rolls (one time in 10,000 you could roll 80 with the sequence 1,1,1,10).
It also leads to the amusing “Yes! Going up!” only to be followed by “Oh, only to 4” moments later.
Ars also has the concepts of simple and quality dice. Simple dice are just a straight 1-10 roll, for un-stressful situations where there is no chance of a catastrophic failure, but no chance of an exceptional success either. Quality dice rolls are the best kind. These rare beasts are for situations where there is no chance of a botch, but a small chance of things going exceptionally well.
The combination of these three basic dice mechanics provides a rich set of options for the Ars Magica GM to call upon to randomise outcomes.
Ars also has lovely rules for long term development. Want to research a level 30 spell but only have a lab total of 32, then that will take you 15 seasons (3.75 years) of downtime†, so it would be better to spend a few seasons studying your arts to bring your lab total up to 38 and then you can complete the research in 4 seasons‡. †2 (32-30) and ‡ 8 (38-30) points per season towards the 30 points required.
If you want, you could allow rolls to be re-rolled with whatever PC luck mechanic you use (Fate points in Warhammer, confidence in Ars, possibilities in Torg etc.). Each re-roll you allow makes it 10x less likely that characters are will botch and gives the players the chance to decide whether they use up that re-roll on the easy, but unlucky roll, or save it for an important roll later.