To clarify: what seems new to me, having been away from D&D for a decade and a half, is the obsession with class balance in the moment. AD&D classes were often balanced over lifecycles rather than just in the moment
You're right, that is (or was, it didn't last long) quite new. In classic D&D, classes and races might 'balance' over the expected life of a campaign, with non-/demi- humans and fighters doing well at low levels, and humans and caster dominating at higher levels. What they discovered when designing 3e, though, was that campaign rarely went that whole life-cycle, they often topped out around 10th. So 3e made some attempt at balance for the first 10 levels, and prettymuch let it go after that. It wasn't really successful in that attempt, but there's a 3.5 variant called 'E6' in which most PC abilities stop advancing at 6th level, that works a bit better.
5E is a new game, by WotC not TSR, in which multiclassing works differently: you pick and choose classes as your total character level increases.
That was a 3e innovation that 5e has adopted, yes.
Lifecycle-oriented balance is something WotC has deliberately eschewed, because they want a fighter 10/wizard 10 to be as much fun to play as a fighter 20 or wizard 20, so the power curve for fighters and wizards is required to be pretty much linear.
That's a good theory. I'm not sure if it holds up, or if 5e uses the 3e model of attempted balance, instead.
5e's immediate predecessor was neatly balanced at all levels, but it didn't use the 3e-style multi-classing or classic Vancian casting that 5e has returned to. If you're seeing a concern for 'balance in the moment,' it's left over from debates and opinions within the community in the 15 years between 2e and 5e. There's little in 5e's design to suggest that it is greatly concerned with balance among classes, either in the moment, or over a life cycle. Rather, if anything, it seeks a roving sort of 'spotlight balance' in which each class is so differentiated and differently specialized that it will, in a well-managed campaign, have moments when it's the star of the show, because it has just the right spell, high skill check, DPR throughput, or whatever to outshine the rest of the party in that given moment.
So it's more differently imbalanced in the moment.
Yeah, this. I get the feeling that DPR obsession is mostly a thing for Internet posters.
Yes. It's easy to calculate, it's reasonably objective.
At the table, it can be fun to be the guy who obtains vital information through spying, or who persuades a key NPC to become an ally, but none of those things can be talked about on the Internet easily because they're all context-dependent. ...
In other words, DPR obsession is a hyperspecialization which is well-adapted for Internet communication and thrives there even moreso than at the table. That's how it looks to me anyway.
DPR optimization has it's place at the table, as well. Since D&D uses a damage-tracking system of hps in which the performance of a figure is not degraded by hp damage until it is dropped/killed, focusing damage on one enemy at a time is the most obvious/basic/effective combat tactic, in general. By choosing a DPR-only class and optimizing it for offense, a player can eschew more complex tactical decisions and non-combat contributions, while still having a character that pulls his own weight and contributes consistently.