The problem with banning or restricting multiclassing or PrCs is that the most powerful people that actively need to be curbed back can be full, straight core, 1-20 clerics, druids, wizards, artificers, etc, etc.
Yes, but viewing PrC's as a solution to the disparity in power (at high levels) between non-casters and full casters is a kludge fix with some awful side-effects.
The PrCs that players actually take can be broken down into three types:
1) Full BAB classes that recieve more than one feat equivalent ability every two levels, and/or better saves, skills lists, and skill points than equivalent base classes.
2) Full caster spell progression classes that recieve more than one feat equivalent ability every five levels, and/or better saves, skills lists, and skill points than equivalent base classes.
3) Classes that all full caster spell progression while simulateously providing most of the benefits of some other class. That is, classes intended to fix the spellcaster multiclassing problem.
Each of these can essentially be boiled down to, "Gives you some large benefit compared to the small cost of entry and the small cost of what you give up." and in particular, "Gives you more feat equivalent class abilities than you could get staying in a base class." Used in this manner, all a PrC amounts to is an admission that full spell progression is worth more than a feat every other level, and that - perforce - either the martial classes or the spells or both are badly designed.
The solution isn't PrCs that provide more to martial classes. The solution is go directly to the problem and make sure you have martial classes that provide you more (and to a certain extent, that spells provide you slightly less). The real and direct solution is making sure that martial classes can, as you put it, "stay afloat". PrCs are a kludge fix to that problem with nasty side-effects. They were never intended to do what they were used to do, and they are bad at it. What they produce is a system that requires a high degree of system mastery and the purchase of a large amount of splat books in order to "stay afloat". They also produce what ultimately turns out to be a very narrow and inflexible system because PrC's have very narrow archetypes and usually very narrow progressions with little flexiblity. And they are unbalancing and typically poorly play-tested, forcing the group to rely primarily on social contract to determine what sort of character is permissible.
Also, it pounds down the horrid idea of class = character that I abhor so much.
PrC's are the ultimate "class = character" concept. The whole idea of a PrC is that it encapsulates not just a set mechanics which can be used to model various heroic powers and abilities, but that it also encapsulates with it certain appearances, loyalties, world-view, and even personality. The write ups of published PrC's aren't generic, but rather character defining. They tell you not only what you can, but who you are. This channels what would be wild creativity into very narrow paths primarily chosen for their mechanical benefit. It doesn't get any more "class = character" than a PrC.
In my game, a concept like 'assassin', or 'thief-catcher', or 'swashbuckler', could be some combination of cleric, wizard, sorcerer, rogue, hunter, fighter, explorer, champion, or shaman (and possibly others). I can't tell from the description and I can imagine viable builds from each class and many combinations. In a game with PrC's, concepts like the above of 'beast master' are literally implemented as 'classes' so that your class is your character, rather than your class(es) being simply an approximation of a set of heroic abilities. In a PrC you are an 'assassin' if you have the assassin class, and there is a strong implication that you can't be a 'real' assassin without one. Now sure, options like rogue or hunter (or some combination) might lend themselves more easily to archetypal 'assassins' just as explorer, fighter, or rogue might lend themselves to 'swashbuckler'.
Originally the idea was that you might enter a PrC from a variaty of different initial points, but in practice PrC's are generally written with one prequisite class clearly in mind. You don't get big books of PrC's where any base class is as likely to multiclass with it as any other. You get splat books written narrowly for particular classes with PrC's written specificly for a particular class that narrow that class definition even further. That is the exact opposite of what you want in true multiclassing, which is that multiclassing broadens the already broad possible archetypes.