A pre-image attack that does not encompass all 64 steps just results in something indistinguishable from the empirical effect of everyone in the world being able to hash faster.
I think the second preimage attack is the one that where we should panic because that means an attacker might be able to rewrite the transaction history.
A first preimage attack just means mining got easier as you said.
No, the attacker cannot rewrite the transaction history because the following reason
support the blockchain is ..., A, B, C, ...
B includes the hash of A, and C includes the hash of B. The preimage attach you mean is to find a B' having a same hash value included in C. However, remember that B' has in the same time to satisfy a lot of constrains: it has to include hash(A), it has to have correct format, and it has to include some transactions beneficial to the attacker
Therefore, to be a successful attack, it is not enough to find a hash(B') == hash(B). You can only modify a small part of B to get a hash(B') == hash(B), that will be much much more difficult than ordinary preimage attack.
