I agree with your #1. Which is why I so strongly disagree with @Ruin Explorer. Infiltration =/= murdering everyone on the way in, getting the Macguffin and then slaughtering the survivors on the way out.
Thanks for shoving words in my mouth buddy, they're delicious and faintly scented from your fingers - been drinking coffee?
I obviously didn't say that. What I've seen happen, repeatedly, is that the PCs manage to get pretty far with little/no violence (and/or only surprise-violence, which D&D admittedly doesn't handle well so that usually does require highly specific approaches or homebrew rules), often retrieving the macguffin(s), then get out, and my experience it's often getting out where things end up "going loud" if they do at all (like a lot of heist media).
But it's heavily in the planning. You have to minimize the opportunities for failure. That means if your plan is to physically do a ton of sneaking past guards, it's a
terrible goddamn plan (this is true with RL heists and many movie heists too, ofc). You want to be using deception, social engineering, magical methods (where not warded against), exploiting weaknesses in security systems, diversions, bribery/extortion, sudden overwhelming force - which doesn't necessarily mean going loud, note, and so on. Like thinking of a specific heist I ran in 4E, which at no point went wrong, I think they made maybe two Stealth checks in the entire thing, and that was only because part of their plan failed. They got in, past all the security systems (which they'd elaborated subverted, particularly by gaining access to the man who originally created them, who didn't even live in the same city as the vault), avoided a bunch of automatons by a couple of clever methods (neither of which I anticipated), and the only fight was inside the vault itself, which like, long story but that wasn't something that could raise an alarm (it wasn't even on the same plane as the rest of the building), then they got out the same way they got in. It was extremely exciting and everyone had a good time. I think you would have too. But this was the result of an entire session and a bit of setup - they prepared a plan, they got all the pieces into place (subverting people, gaining access to devices, getting hold of plans, the classic "posing as a client" scouting method, and so on), and only then did they execute.
You seem to be thinking of a heist as like, "Here's a building with stuff in it, you're going to need to sneak past a bunch of guards, steal the stuff, and sneak out", and it's like, no. That's a terrible plan. That's not even a plan. A heist is when you prep elaborately and get the pieces into place, and when you have fallbacks and so on. Yeah it might go wrong, but if early elements of the plan rely on physically sneaking past people and if that fails the whole plan is blown? That is unacceptably bad plan. This is like, the problem I see with your criticism, like if a plan has a single point of failure like that, like, where one role really could blow the whole thing - that's a bad plan! Call it off! Find a different approach. If one roll increases the risk, or makes it harder or might cost resources or the like, sure that might be a viable plan, but one roll = total chaos breaks out, well, that roll better come after you got the macguffin in the plan!
It does help that my main group are extremely smart, lateral thinkers who
like thinking and planning, are familiar with this sort of genre, and who very much want to be Locke Lamora or Kaz Brekker and crew, particularly enjoy masquerading as people, and like actually want to come up with the plan. I've definitely played in groups that were huge fun but absolutely
could not have managed any kind of situation like this, because they were just living for the drama, and BitD would definitely be how I'd do a heist with them (or at least D&D with more modifications). I'd also say that as teenagers, the same group were
not good at this sort of thing. In SR and CP2020, countless heists went tits-up because of fundamentally bad plans.
I do think D&D's one linear die approach to skills, plus the really limited use of Inspiration (which just gives you Advantage, which you may well have anyway), in the default RAW, does really limit things. In something like SR, heists aren't fundamentally better supported than D&D by the terms you've been using (because there are no specific mechanics relating to them), but it uses a multi-die system, and in various editions there have been ways to re-roll things or boost vital rolls significantly (moreso than Advantage does in D&D), so in SR you can rely on plans that are a lot riskier than D&D ones, a lot less fundamentally planned-through. I think this one-die thing, together with the lack of rules specifically for KOing/killing people from surprise (which some D&D-relatives like WWN/SWN do have), are the biggest cramps on heists in D&D, because they force you to use more elaborate plans and to prep a lot more, and tend to make magic more important (though in 4E they seemed to be fine with only Rituals). WWN/SWN solves both those problems note, so I'd expect it to be significantly easier to run "looser" heists with either.
Heck, just the notion of what a failed check means is hard to reproduce as how one DM interprets failure can be radically different from another.
Sure, and this is one thing that PtbA games have over almost all other RPGs on the market, in that they do more to attempt to define how a failure can be interpreted by the DM than anything else I can think of. Is that a strike against D&D or a strike against most non-PtbA RPGs though? I'd also repeat that any plan that relies on a single check succeeding, particularly in a system like D&D, is a bad or at least desperate plan. That doesn't just applies to heists, of course, either. D&D is particularly hard to work with failure-wise because not only do DMs potentially interpret failure widely (as with a lot of RPGs), but success/failure is binary, unless the optional rules on page 242 of the DMG are in play - and even if they are, they still make D&D inelegant (D&D not being designed for margin of success/failure or success-at-a-cost) and you still have the "one die" problem to some extent.
Of course even PtbA games are somewhat subject to this - some DMs with those are far more or far less aggressive with the moves they make with failures or success-at-a-cost where the cost isn't precisely specified. Or are just a broken record on what moves they make.