So, I was thinking about how I almost never prestige with any character, and I think it comes down to this: Prestige Classes as they are have a bunch of design philosophies, and most of them are terrible. Well, and that it's just hard to search through them because they're not sorted by what character can even think of taking them.
- Prestige Classes Should Be Weirdly Over-Specific: By RaW, you're supposed to make up a Prestige Class for your character in the middle of the campaign based on what happened. So you'd spend a year on the river Styx, and then you'd write a Boatman of Styx prestige class that requires you to do that, which conveniently you happen to qualify for. It's an amusing idea, but any prestige class written and posted according to this philosophy inherently violates this philosophy. Except Elothar, because it's a joke.
- Prestige Classes Should be an Awkward Patch for Multiclassing: Like a third of all Prestige Classes are just "class type A and class type B multiclass poorly, what if we made a class that advanced both core features but set them back 3 levels in each?" And then if you actually try to take them you find a way to get set back only one or two levels in each, because there usually is one. At the point where you're thinking of writing a bunch of prestige classes like this for your class, maybe you should step back and just say "this class advances its [class feature] according to its level -3 when multiclassing" or something like that.
- Prestige Classes Should Let You Expand Your Character Concept in Ways That Aren't Appropriate for Level 1: This sounds like a nice and sensible idea, but there was another Tome design philosophy: "Your character concept should be playable from level 1". That philosophy won, almost every character concept that is available as a prestige class is also available as a base class at this point.
- Prestige Classes Should be Pseudo-Gestalt for Spellcasters: This one is fine actually. Basically you give a set of side features that are a bit too involved to be a feat or feat tree on top of your spellcasting, and spellcasters are kind of written with this in mind, and it works out more or less.
- Allowing Late-Entry To Classes That Aren't Multiclass-Friendly: Base classes are generally written to give level 1 abilities at level 1. For fighter-types it actually sometimes works fine to just multiclass, because fighter-type abilities are often multiplicative (boosting actions you can already do) rather than additive (giving you an exponentially more powerful alternative use of your action) like a wizard. A 1 level dip into Kineticist continues to give level appropriate benefits to a level 11 character. Spheredipping works fine too. But what if you want something that isn't? Of course, the problem with this is that if the PrC is written to be level-appropriate, then you'd need to write an alternate version for level 7 entry vs level 9 entry...
- Integrating Alternate Abilities Into an Existing Class: Sometimes people write classes with unique and interesting resource management. Naturally this doesn't multiclass well, because even if you could find another class that was powerful enough to combine it with, it would be a pain to manage two resource systems at once. Potentially one could write a PrC that added necromantic powers to the Elemental Siphon's axes or let a Crusader cast Divine spells with their maneuver slots to let the player grow and change their character's concept without overconvoluting how they play. Of course, the problem with this is that if you write the resource system you could just write multiclassing-friendly rules for multiclassing with other classes that use the same resource system, and then write all the PrCs as base classes that use that system instead...