I totally understand your reasoning, and concede that it is 100% logical, but I disagree for one major reason. D&D isn't a simulation of the real world. If I can wrap my head around giant flying reptiles that breathe pure cold, giants so large their legs should shatter, hippies turning into dire squirrels, nerds casting spells, and enraged half-wits shrugging off blows that should maim them instantly, what is wrong with shady pickpockets being unnaturally good at shooting you when you are looking away for a split second?
In your opinion, would it be reasonable to believe that rogues have an uncanny ability to time their strikes for maximum effect that is unique to their class? They seem to be capable of doing so when a target is engaged with another foe or even distracted by a familiar, or at least I haven't seen any arguments to the contrary. Making the leap to include peeking out and attacking from even poor concealment seems perfectly natural to me. It doesn't really matter that you know where the shot is coming from if you don't know when it is coming.
Honestly, in OP's example, I could see granting the sneak attack but not making the attack at advantage. It's not RAW, but I think it's a reasonable compromise at any table having this debate. For most of us, we seem to have a consensus at our tables one way or another. Both work as RAW with different interpretations of how to apply the rules, and its another example of where the designers failed to be clear enough to prevent dozens of pages of arguments that have in several instances passed beyond civil discourse.
First, I agree that D&D is not a particularly realistic simulation. I do think that a rogue should get advantage in combat at least occasionally, or even on multiple turns in a combat with enough chaos and cover. It's not that they would
never get to attack after being hidden, it's a question of is the target anticipating an attack from that location for me.
If you have ... say a hundred archer goblins all firing at you at once but they weren't hidden before the attack they would not get advantage, correct? Because the assumption is that while you didn't know exactly who in your party they would be attacking, they were there and they were going to attack at some point. Your PC was anticipating the attack, even if the goblins were all firing from different directions they had no advantage. So I don't think it's a stretch to say that
if you know where a rogue was attacking from, you are also anticipating their attack and there would be no advantage.
One simple example: there are guards in the hall that are alert and ready for trouble. Typical, 10 foot wide corridor with a T-intersection that they're staring at. If a rogue comes around the corner they're staring at, I'm sorry but the rogue is not going to get advantage. They see you coming. On the other hand, if the rogue can stealthily get behind the guards to attack they will get advantage; the guards were expecting the attack from the T intersection in front of them, not the open door behind them.
If you want to rule differently that's fine as well. There's just too many situations to have a blanket rule without DM adjudication and some level of "does this at least kind-of-sort-of make sense to me". There are simply too many variables, too many preferences, to come up with hard-and-fast rules. Rules in previous editions often left our group scratching our heads thinking "how the **** would that work" because it was just too mechanical, too much rules over rulings. For us, it felt like a board game.
So I have no problem with how people run it. I just have a problem with people quoting 1 sentence without considering anything else the rules say means "RAW says I'm right" or saying that "this is the correct way to run it because it makes sense to me".