Caveat the First: I Am Making Broad Generalizations
There are cases where you might want to break these.
Caveat the Second: I Am Not Overly Familiar With Pathfinder
I have read the Core rule book, but none of the supplemental material, including the archetypes (alternate class features in 3.5 lingo), and Paizo did make it a goal to counter, at the very least, my point #2. Within the Core material that I have read, I would argue that they largely failed. I have heard that they have done somewhat better with the supplemental material, but I would not know. Mostly, the answer is really only being given for 3.5, with the understanding that Pathfinder Core changed very little in any significant way.
Caveat the Third: I View Classes as Purely Metagame Concepts
I have little to no problem with multiclassing, even a great deal of multiclassing, because in my games a character does not think of himself as a Paladin 6 or a Barbarian 2/Rogue 1/Cleric 1/Swordsage 2 – he thinks of himself as a knight, or a mercenary, or a mystic, or whatever.
Or, say, a Monk 2/Paladin 6 thinks of herself as a Samurai.
Each class is, as far as I am concerned, a bundle of mechanical features with some suggested ideas for what sort of person would have them. Players at my table know, for instance, that they don’t even have to ask if they want to treat, for instance, Rage as instead “Zen Focus,” so long as the mechanics do not change. Most abilities are reasonably generic and can be understood as a lot of possible things in character.
Multiclass characters can represent a change of heart – someone who used to train one way abandoning it for another – but particularly when you start at levels above 1 and you start with levels in a few classes, it often makes more sense to think of a multiclassed character as someone who pursued one path their whole life – that path is just mechanically represented by different classes.
1. Spellcasters should not multiclass
Spellcasters, as well as manifesters, will always benefit most from getting the highest-level spells (powers) available at a given level. Their spellcasting (manifesting) progressions benefit only from more (effective) levels in their class.
Note that technically in 3.5 jargon, prestige classes do not count as multiclassing, but a separate thing. Prestige classes can be very good for spellcasters, since they frequently have few or no class features aside from their spellcasting, and many prestige classes advance spellcasting on a 1:1 basis. In some cases, this can literally be “something for nothing,” as the prestige class provides class features while not costing a spellcaster anything from their base class.
2. Mundanes should multiclass
Mundane classes (and the weaker half-casters, usually the ones who stop at 4th-level spells) do not have a solid progression like spellcasters do. Furthermore, most of them are extremely front-loaded. For examples:
- Barbarian gives Rage, and potentially Pounce (Lion Spirit Totem, Complete Champion) in one level. Level 2 can get you Improved Trip without prerequisites (Wolf Totem, Unearthed Arcana). Beyond that, the next significant benefit comes at 11th level (Greater Rage).
- Fighter gives a lot of proficiencies and a feat at 1st level, and another feat at 2nd, but nothing at third, and thereafter the rate of ½ feats per level, which is not good.
- Monk gives a couple of feats in addition to improved Improved Unarmed Strike, plus Evasion and Flurry of Blows, in the first two levels. It proceeds to give nothing much for the next 18.
- Ninja (Complete Adventurer) gets Ghost Step at 2nd, and little else until it gains the ability to become Ethereal.
- Ranger can get some combat feats without prerequisites and gets a feat per level until 3rd, but is overall not a great class.
- Rogue, and most Sneak Attack-ing classes, grant 1d6 Sneak Attack damage each odd level. By taking an odd number of levels in multiple Sneak Attack-granting classes, you can gain higher Sneak Attack damage than a single-classed Rogue would get.
And so on. So a Barbarian 2/Fighter 2/Monk 2/Paladin 2 has way more class features than a Barbarian 8, Ranger 8, Fighter 8, or Paladin 8.
Note that Cleric 1 is quite probably the best single-level dip in the game, despite also being a fullcaster that you can focus all 20 levels on. In this sense, Clerics are both martial and magical. Even characters without enough Wisdom to cast Cleric spells can make good use of a Cleric dip.
Psychic Warriors are another exception: if you aren’t absolutely requiring full BAB, Psychic Warrior 2 can get you the same feats that Fighter 2 can, plus a few Powers which can be very useful to an otherwise-mundane warrior.
3. Those in between can go either way or, sometimes, halfway between
You can dip Bard for Bardic Knowledge, Inspire Courage, and fascinate, or you could focus on Bard to get its quality spellcasting. Binders (Tome of Magic) and meldshapers (Magic of Incarnum) can be dipped for a select Vestige or Chakra bind, or focused on to maximize those features.
Factota (Dungeonscape) go even further and are solid at 1 (all skills in class), 3 (Brains over Brawn), 8 (Cunning Surge), or 20 levels. Binders and meldshapers are also reasonably good at being worth however many levels of them you want to take.
4. Tome of Battle is exceptional and unique
Tome of Battle classes multiclass better than any other classes in the system, because they add half their level in other classes to their Initiator Level, and can select higher-level maneuvers based on this improved Initiator Level. That means unlike a Fighter 8/Wizard 1, who gets 1st-level spells, a Fighter 8/Warblade 1 has Initiator Level 5 and gets 3rd-level maneuvers.
Ardents (Complete Psionic) have a somewhat-similar mechanic for determining what level of power they can use, but no built-in bonuses to their Manifester Level from other classes. The Practiced Manifester feat (also Complete Psionic), however, means that an Ardent can take up to 4 levels in other classes while maintaining full Manifester Level and therefore the highest-level powers.
You're in quite a difficult situation. Your players don't have information, don't have many leads, have one dead party member, and have been launched into confusion.
Slow down the overarching plot of your game - grind it to a halt for now, if you need to. Your players (and their characters) both are not ready for it and do not have the information they need to process it. The way this game is being played, you have significant control over the plot and pacing of the game. Slow, possibly to a halt, the destruction of the universe and the whole ancient magicks and dragons shindig. Your players know it's an imminent threat, but lack the skills or knowledge they need to confront it in any way - they need time to learn and grow.
In a lot of ways, this can serve simply to set the scene for whatever else is going on, and contextualize the additional information the players gain from other goals and quests. Let it do so - your players will learn about the overall problem, and potentially be able to connect the dots where others have not. At a later time, once the players have better information, they'll be able to handle the overarching plot. Until then...
Give your players information. Consistently, the examples you've communicated have had a common problem: the players achieve what they in the scene want, but don't get any information about what they want in the broader scope. There are two effects from this: first, the players will grow frustrated; second: the players will quickly run out of leads. The leads they get need to go somewhere.
Maybe occasionally you can throw in some lead that just doesn't work, but the majority of them should provide some sense of success. The players learn something, or gain something. They don't have to completely understand it, but something needs to happen. Otherwise, all the plot hooks will drop. For instance, while interrogating the cultists, some design element of this scene needs to provide the characters with more information about who and what they want to be pursuing.
Always ask yourself, "Would the players really enjoy this right now?" - and give a more detailed answer than just yes or no. For instance, concerning the meteor falling, ask yourself, "Would the players enjoy the meteor falling right now?" The answer to this question is, as you've retrospectively identified, something along the lines of "No, because the players do not have the information or skill they need to handle the situation."
You've got a good scene - meteor falls, chaos ensues - but the players aren't prepared to handle it in a way that makes the game fun for them. If the game were run again, this scene should be saved for a time when the players can deal with it in a meaningful way. By answering the "is it fun?" question, you've learned two areas which must be fulfilled before you can safely play out this scene: the players need to be more powerful, and the players need more information. Once the players have achieved this, then you can go and run this scene.
Obviously, this scene has already been run, and I'm only using it as an example. Short of rolling the game back to before this scene (which may be a completely reasonable course of action, depending), this is part of the game world right now, and is just something the players are going to have to deal with.
This is the core of my advice. But how does it apply to your campaign?
Here's what I'd strongly advise: set aside things the players can't handle for now. They can evolve in the background, but as a general rule, they shouldn't affect the players until they're ready to be drawn into those plot threads.
Come up with some scenes that your players would all want to play, and would all be able to handle, and draw the characters into them. Your players will decide what they want to do next, and it's probably going to be to explore deeper. Through these scenes, you can feed them pieces of information, ideas, thoughts, and leads, which will draw them both a) into more power through the experience system, and b) into greater understanding of the forces at work in your game.
Best Answer
Just upfront, the reality is that the primary answer to this question is that almost everything in the game with LA has way too much LA. It seems that the designers balanced LA against pretty much the lowest common denominator of what HD might grant—assuming that they really tried to balance it at all. In a lot of cases, it seems more likely that they just eyeballed it and erred on the side of caution by putting more LA.
But that isn’t precisely what was asked: the question is more like, why is LA crippling even if you get a lot out of it?
Every HD comes with some basic things: hp, including the Con bonus, skill points and the increase in your maximum skill rank, and increases to base saving throw bonuses. On top of that we have feats, BAB, spellcasting, and so on that don’t necessarily happen every HD but are generally part of your HD progression. In short, nearly all forms of character growth.
LA means you don’t get any of that, and things that come with LA generally don’t replace any of that.
Whatever you get for LA, it probably isn’t going to be the kinds of things you’d get from HD. Even if you get really good stuff, it’s going to be good stuff that’s orthogonal to the above things that are, effectively, your growth potential. That is, LA is giving you some random thing that can’t and won’t grow, and in exchange it slows your growth. So even if it was a good trade at the time, it quickly becomes less of one—as noted by Unearthed Arcana when introducing the buy-off rules.
Also, not for nothing: those character growth stats that everyone gets from their HD are the primary things that numerically stand between you and death. Even with Con bonuses, LA’d characters often have very little hp, and poor saves. Fundamentally, even “good trades” in LA still cause enormous problems due to character skew—being too-good at some things and too-poor at others. This causes fundamental and often insurmountable difficulties in creating appropriate challenges for the whole party.
So does buy-off solve the problem? No, or at least not quickly, particularly at higher amounts of LA. LA +1 is bought off at 3rd, LA +2 is bought off at 6th and 9th, LA +3 is bought off at 9th, 15th, and 18th, and anything above that can’t be bought off entirely without going into epic (which Unearthed Arcana does not provide buy-off rules for). Importantly, even when you buy off an LA, nothing immediately changes—you are still without that level, because of the XP cost of the buy-off. You just start gaining more XP so that you can eventually catch up. So LA +1 costs 3,000 XP total, LA +2 costs 16.000 XP total, and LA +3 costs 45,000 XP total. That is a lot of XP, and while you will recoup those losses due to the greater XP you earn as a lower-level character, it won’t happen all that quickly. The 3,000 XP for LA +1 isn’t too bad, and you can start working your way back at 3rd, but you can’t actually start to catch up with your allies after paying off LA +2 until 9th (which may itself happen later for you than for your allies), and for LA +3 it’s 18th (and likewise that might itself be delayed). And again, this is something that is directly standing between you and the growth of some of the most significant numbers that are keeping you alive.
If you’re starting at 20th, then whatever. But most people don’t do that—and generally speaking, you shouldn’t, because D&D 3.5e is very-nearly guaranteed to be non-functional at that point. Holding out for that buy-off you’ll be able to do someday is not likely to be an enjoyable experience, because odds are good you’ll never get there (either cuz the character dies or the campaign does), and even if you eventually do it’ll be a slog to get there, and it’s just not worth it. Again, LA +1, not so bad, with buy-off, but it gets a lot worse as LA goes up.
But it has to be repeated that this argument is made much easier by the near-complete lack of compelling LA +2, LA +3 options that could maybe cause us to look at the drawbacks a little bit harder. As things stand, it’s easy to say the drawbacks are substantial (again, barring games started at very-high levels), and the benefits are mediocre, and be content to leave our analysis there. But I stand by the argument even for high-power templates like saint, again because of the skew that it creates in a character.