Setting a block size limit of 1MB was, and continues to be a hacky workaround.
It is certainly not a hacky workaround. It is a limit that was needed (it still is for the time being).
Theory drives development, but in practice sometimes hacky workarounds are needed.
If it can be avoided, not really.
The block size limit was a hacky workaround to the expensive to validate issue. An issue that is now mitigated by other much better solutions, not least a well incentivised distributed mining economy. That is now smart enough to route around such an attack, making it prohibitively expensive to maintain.
So exactly what is the plan, replace one "hacky workaround" with another? Quite a lovely way forward. Segwit is being delivered and it will ease the validation problem and increase the transaction capacity. What is the problem exactly?
Problem: an attacker can create a block that is so expensive to validate that other miners would get stuck validation the block.
Hack: Set an arbitrary limit which is way above what we need right now, but closes the attack vector.
Solution: 1 transaction blocks.
Problem: the block size limit is causing transactions to get stuck in the mempool
Hack: raise the block size limit to 2MB
Solution: remove the block size limit
Segwit isn't a solution designed to fix the block size limit. Its a solution to another problem that right now is undefined, that is being sold as a solution to a problem that is being actively curated by those who refuse to remove a prior temporary hack.
What problem is it that requires signatures to be segregated into another data structure and not counted towards the fees. Nobody can give a straight answer to that very simple question. Why is witness data priced differently?