Increasing max block size is not what "scaling" means. If you can process M transactions, and it takes N bytes, then processing 1000*M transactions by using 1000*N bytes is not "scaling". It is the simplest "linear growth" and has nothing to do with real scaling. The "real scaling" is called "compression". For example, where you can handle N-of-N multisig by using a single signature, instead of N signatures. Or where you can compress a chain of A->B->C->...->Z transactions into A->Z transaction, that is also scaling. And that is the right direction for the next soft-forks.

I think it is gone. Also because BCH and similar chains did it in a hard-fork way for no reason. They could create 1 GB blocks in a soft-fork way, and we can also do that, if it will turn out that no other way is possible. But as far as I know, there are better ways to scale things, so I think we won't need increasing the block size in the future. Another thing is that the max witness size is just a parameter that any node can set to anything, so from the code perspective, everything is ready, it is just a matter of reaching consensus if needed.