Mutliclassing in 5e is a half baked optional rule, the land of broken (mostly very underpowered) combos. One good reason for Larian to avoid it is it gives players without 5e system mastery a whole lot of trap options to fall into.
for example, while Wizard/Cleric combine very poorly by cutting off access to high level spells.
Actually, if you include wizard, it's not so bad. Caster levels stack for purpose of spell slot allocation, but they don't for spells known/prepared. But wizards can still learn the higher level spells from scrolls. It's things like cleric/druid that are completely locked out of the higher level spells. But to implement separate list of spells known/prepared for multiclassing would require a compete do-over of the BG3 character sheet. And you need to find room for up to 9 separate spell lists. Unless you cap the number of times you can multiclass of course. But that's not PHB.
And Larian have created additional problems for multiclassing in the way they have implemented certain features. For example "Paladin Oath Charge" should be Channel Divinity, and share a pool of uses in a cleric/paladin multiclass. The way Sneak Attack is implemented would not interact correctly with Extra Attack in a Fighter/Rogue, and Barbarian/Draconic Sorcerer armor class would be completely borked.
But it is also a very popular feature that Larian have stated they were going to add.
It's easy to promise things before you have fully assessed how challenging they are going to be. As a project manager, people making the client ridiculous promises is the story of my life!
Definitely ways this could be broken or overpowered, though I hardly think Larian considers that a bad thing. If anything, the fact that they haven't addressed things like BA shove and hide for everyone suggests to me that they are fine having brokenly powerful combos in the game, even if that comes at the cost of certain characters. Rogues lose a lot of value since everyone can BA hide, for example. Besides, even with an extra charge of Channel Divinity, Paladin/Cleric would be far short of multiclasses like Paladin/Sorc.
As for Sneak Attack and Unarmored Defense, they shouldn't be too problematic. Just have Sneak Attack count as an attack action for the purpose of Extra Attack (which the homebrew weapon abilities already do, so the coding needed should already exist). And just make it so that the unarmored defense and draconic defense don't stack, but takes the highest value.
The problem will definitely be prepared spellcasters classing with other prepared spellcasters, though the fact that Wizards were able to learn spells from other classes from scrolls in BG3 suggests they might've been working on something that included mixing together spell lists.
Until otherwise stated, I'll take their promise of multiclassing to mean that they are going to add multiclassing in some form, though it might differ from the tabletop version in some key ways. They might come out and say they couldn't get it implemented later, which wouldn't surprise me, but I don't think they have yet.
