I think that's an interesting line of discussion.
That being said, the one glaring difference between knowing the results of an attack roll versus a sensory based skill like Perception or Insight is direct, observable results in the case of failure.
If I swing a sword at someone as part of what would normally be an excellent attack (a high roll) yet I still miss, I know my target is either very nimble or heavily armored or protected by some other means evident via the DM's narration of the attack roll result. If I fail to perceive a possible threat via a Perception check, I have no feedback by which to judge my effort.
The lack of feedback in sensory-based checks is another reason I believe passive scores are an excellent tool. If nothing else, they provide a consistent standard as a reliable data point.
I am not talking about knowing success or fail - right?
look at the examples...
To-hit rolls:
Roll a 2+9=11 miss - player makes no in-character changes to tactics.
Roll a 19+9=28 miss - player goes "oh crap" and makes in character changes to tactics like swithcing to saves spells instead of attack spells or switching to help or grapple or advantage.
In both cases they got immediate feedback that said fail.
In both cases, they made judgements about how to proceed based on in part knowing the D20 roll that yielded the result.
Mostly likely the GM narrated those two results differently - one is a wet spot slip, the other is a dead on swing that stilll failed due to... insert Gm narration.
Now look at the example provided in the response - move silently.
If i make a stealth check why is it wrong for me to read the roll of 2 as "i sucked at that and made noise" and assume i have been heard (right or wrong) but to read the 19 roll as "i nailed that likely almost totally silent" and act like i am still hidden (right or wrong)? Reference many times in various source we see twig snaps, stairs creek, etc.
Why does a Gm choose to not narrate the 2 and the 19 differently on the stealth check to provide the player/character the same type of feedback?
Look to Insight then. You are looking for clues.
roll 2: There are a lot of mixed signals and its hard to hear his inflection over the nosie and everybody is getting jostled in the crowd so its hard to be sure what is reacting to what - but you do not see any clear sign he is lying. (or what he is planning.)
roll 19: Clear view of his face, easy to hear when he directly responds and clear tells all add up to show you with no exceptions that he seems to be being truthful. there are no signs of deception.
Now in either case, maybe you are right, maybe you are wrong. You do not know the opposing DC - but the former case leaves you very much doubting that you can trust the outcome and the latter cases leaves you thinking that unless he is really really good you can trust the result.
Why as Gm choose to narrate the result of that check and its d20 roll in a way to "hide" the d20?
Picking locks, moving silently, searching a room - all can be narrated in a way that reflects the d20 just like we reflect the "quality of the roll" for an attack (or at least we allow it to be assumed the player/character can work off that roll to make decisions.)
or do you play in such a way that the swing roll of 2 and 19 have to be treated as unknowns and tactics not change based on them.
"the roll" is not to me about success/failure but about the myriad intangibles we work into the game scene after scene - but for some - only for some d20 rolls.
In my games, all d20 rolls are treated as "known" and reflecting "confidence of the quality of the effort" whether its a swing or a search or an insight or move silently. i provide narration when needed to support that.
The i do not run into worries of who rolls what and when they can use the roll feedback and when they cannot and things run very smoothly and consistently.