Given that they’ve developed a faithful and fairly wide-ranging representation of D&D 5e, I’m willing to bet that ended up being a lot more involved than their own proprietary system.

That game was just one example, but since you seem interested in singling it out:

Turn-based game rules cannot explain the awful graphics performance that game has, even at idle, on some systems. (Not even D&D 5e, which I happen to know in detail.)

Graphics engine enhancements might explain it, but in that case, the developers should have included options to disable those enhancements.

I haven't reverse engineered the code, but some of the behaviors I've seen in that game smell strongly of decisions/mistakes that I would expect from a game that was rushed, such as lack of occlusion culling. Others smell like mistakes that are common among programmers who haven't yet learned how to use the graphics APIs efficiently, such as rapid-fire operations that should instead be batched. Still others could be explained by poor texture and/or model scaling techniques. As a software engineer, the bad performance in this particular game looks like it could come from a combination of several different factors. None of them are new in this field. All of them can usually be avoided or mitigated.

In any case, the point is that none of that analysis matters for the sake of this discussion, because a community with experience using products doesn't have to be experienced in building them in order to notice when something is wrong. It's not fair to categorically dismiss their criticism.

(Thankfully, the Baldur's Gate 3 developers haven't dismissed it. Instead, they are working on improving it. Better late than never.)