Stealth is not a few tweaks away from great; it’s an overhaul away from passable.
I'm stealing this sentence and putting it on a plaque somewhere.
It occurs to me that all 5 of OP’s proposed changes taken in conjunction would make stealthing very difficult. Not that that’s a bad thing, but the impact on gameplay deserves examination. What better way than by playtesting?
I proposed changes that I thought would bring BG3's stealth mechanics in line with DnD 5e. Not that 5e is some holy grail of good rpg design (far from it), but I believe its stealth system works reasonably well and would be a safe baseline for Larian to implement. In the years I've DM'd 5e my experience is that it's easy for a specialized character (e.g Rogue or Ranger with Expertise) to achieve a +10 Stealth bonus by level 5, which is absolutely enough to consistently avoid notice as most published creatures have a Passive Perception around 13, often less. And that's not accounting for the Skulker feat, spells like Pass without trace, magic items like the cloak and boots of elvenkind and other stealth-improving options that can turn even a fighter with moderate Dexterity into a decent sneak.
TL;DR I think RAW 5e stealth would work fine with little playtesting needed and leave Stealth and ambushes as a rewarding option for characters who choose to specialize in it.
I just read the recent post about Shove apparently still being a bonus action 20-foot push. It's quite worrisome and it implies to me that Larian don't care strongly about mechanical balance and fairness, which doesn't bode well for the stealth system also. I'm ready to be surprised, but I won't hold my breath.
