I'm not arguing for a centrally defined salary committee, but surely you have to wonder if there is a better way?
This has been well studied. The answer is that compensation is proportional to marginal productivity rather than mean productivity.
I.e., consider a very important profession, say superhero. Having one superhero to stop the Joker is worth $1M. Having two superheros is worth $1.1M - Batman is sufficient to stop the very dangerous Joker, but two superheros can stop both the Joker and the far less dangerous Harleyquinn. The marginal productivity of an extra superhero is $100k even if the mean productivity is $550k. As a result, superheros get paid $100k because if any superhero quits, society loses only $100k rather than $550k.
An extra banker produces a lot of value. An extra teacher produces very little. Hence, teacher pay is lower than banker pay.
(As a proponent of the signalling theory of education, I'd also argue that the value of a lot of education is actually negative.)
For some value of "value".
(Also, the example of two superheros isn't a good one - they will probably collude and get something much closer to their full value)
It's also the case that marginal arguments are very clear when the examples use small numbers. It has the side effect of making them sound silly. But I guess there are not 2 of very many real professions.
That seems to be the most reasonable argument for redistributive taxation - the idea that the >$1M earner can afford to part with some of his surplus to fund programs that benefit society in its entirety, and more specifically those whose jobs are not so richly compensated.
Of course, the devil is in the details and implementation.