Insight Played Wrong Can Become The Most Unrealistic Thing In ...

I think there are a ton of good points put forward here.

One thing I’ve started to think lately as a DM is that all opposed checks should be vs passive NPC/monster scores. The passive stat is the ‘average’ you’d get over many rolls right? Why throw more random chance into the mix. The player rolling d20 on top of a (probably 20% bonus at the most commonly played levels) is already random enough.

For this reason I’m also going to start trying 3 checks for most things. I view these ‘encounters’ as alternatives to combat ‘encounters’ which are basically a string of Str or Dex checks (or whatever combat stat is used). Combat encounters are seldom determined by a single roll, why should others be? If it’s not worth the 3 (or more) rolls and associated narrative, maybe skip the checks? This is what I plan to explore anyway (along with replacing Wisdom with Awareness, as in 5e its clearly more about perception (spotting and hearing stuff etc).

With 3 checks you get more nuance in terms of failure and success, and less chance of one bad roll spoiling things. Also, if it’s vital info for the plot progression, don’t base it on a check like this or you have to railroad around it. A single roll might be fine for something like a really fast (6 second) scan of room to spot something, because that is kinda/almost random. If they spend a full minute or more looking over the room and the thing is spottable, they should spot it (as in take 10 or take 20 rules). If they carefully search a room and a clue/thing is hidden somewhere, they should find it, otherwise why bother having it there? One caveat to that is a scenario where they are likely to search several rooms. You can have the DC go from high in the first room down to 1 in the last room. This works fine with narrative.

Side note (personal peeve) – if you adjust DC’s based on party skill you’ve thrown out the whole system. Now you’re just making arbitrary calls on the odds of success and are punishing players for investing in stats/skills.

Other side note, the group spotting an liar problem also works in reverse (both ways really) with stealth checks. A rouge can’t effectively stealth or hide against a group of 4 of more monsters rolling individually. A party can’t stealth against 1 creature if they all have to roll (well, the odds are terrible). Even a group check against an opposed roll is just super random. Letting the sneaker, or sneaking group take 10 against a passive perception would be better IMHO.