Think of these as things which become critical after the engineer has mastered the skill of solving a hard problem.
There's only so much you can improve if the output is low. After all, if all your seniors spend their days managing, who does the actual coding? Juniors. And what happens with them after they've "mastered the skill of solving a hard problem"? They become seniors, stop coding and start teaching the next batch of juniors.
This seems like purposefully limiting the quality of the work, and increasing the time it takes.
> there becomes a point of diminishing returns in terms of pure engineering skill
It's quite surprising, if we think about it - software engineering is about as close as you can get to pure productivity multiplier. It shouldn't have diminishing returns, not so soon. It should have exponential returns. After all, the same skills that can be applied to a problem, can be also applied to a meta-problem of solving the problem faster.
By that town, all managers need to go do some technical work and get better at that.