Well you can't reverse sha256 to get the same message back, which means even if you manage to reverse it and find a message that hashes to the same output hash, you have just found a hash collision, because the original message could be anything.
Not sure what you're trying to say here. The thing is: there is no point to talking about "reversal", because it is undefined. Hash functions are simply one-way. There may be an algorithm some day that takes as input a hash and produces as output one pre-image. But, there are theoretically more than that. It is the same as talking about "reversal" to a modulus operation.
With current hardware, to even have a theoretical chance at ECDLP, one needs more than 28*1015 qubits
What is the reasoning behind that number?