I am inclined to develop super-powered campaigns, because dungeon crawl RPGs are tainted with some widespread cultural conventions.
1. If you have magic, make sure that it can run out of ammo.
2. If you have magic, start the wizard with a max of one good spell. Make him blow his load early, and don't give him a chance to recover his magic so that he can be sidelined for most of the play session.
3. if the wizard wants more than one spell, give them out only at higher levels. That way, the wizard's showboating is dependent on the group getting together every week for a few months so that people can level up.
It gets to the point where every player character has to start out on par with the X-Men or else there will not be anything but boring dungeon-crawling, trap checking, and battles of attrition.
Of course, magic is not interesting unless there are things that magic cannot do, and that applies to superpowers as well. Superman is (IMHO) much more boring than Batman because Superman can skip to the most efficient answer - heat vision, frost breath, flying, super-strength - and he never gets hurt unless you whack him with Kryptonite. Batman, on the other hand, never seems to run out of stamina, willpower, dirty tricks, and hidden knowledge. Batman's plot armor is probably even more powerful than Superman's invincibility. The thing about Batman is that the writers pretend that his super-luck, super-skill, and super-everything has some grounding in physical reality. Thus Batman stories have to be written with the theme that Batman's gimmicks are not supernatural. Reading the fantasy of super-science is the interesting part.
In practical RPG terms, Ars Magica does something similar with its Aristotelian pseudo-science. Ars Magica doesn't allow players to skip to the most efficient solution - they have to justify it in terms of misspelled Latin pseudo-Aristotelian science. However, Ars Magica does suggest that magic costs stamina, so you do run out of spells every day - but the game is paced on seasons, so a day of spell-casting is a small part of playtime.
Mage: the Ascension might be a workable way to do this. Unfortunately, even though I *think* that I got Mage to work perfectly circa 1993, the White Wolf playerbase is more cancerous than most groups of gamers, and I don't know whether my Mage storytelling sucked or was awesome or both sucked and was awesome at the same time. The advantage of Mage was that I could recruit lots of players because White Wolf was popular in 1993.
Now, I could preach that GURPS Thaumatology is the perfect set of magic rules, but that is only half-true. GURPS Thaumatology is a rule creation kit. It is not a set of playable rules, it is a set of guidelines that can show the GM how to design a set of playable rules. To make the game work, you need a theme. Furthermore, you need extensive playtesting, and I cannot recruit many GURPS players.
So here I am - I think that I know how to design magic rules that allow players to cast unlimited spells without ever running out of ammo. I think that my GM experience has allowed me to transcend the painful AD&D failures that shaped me. But I can only run games rarely, for small groups, so I will never know whether my designs are really good or just fantasy heartbreakers.