In PoW your likelihood of mining a block is independent of the last hash you made, so incrementing the nonce, replacing a tx, or replacing all txs before you hash the header again, all have the same odds of finding the block.

Ah, I see. Thanks for confirming that this is indeed the case. I wasn't so sure even though I read something similar already.

Probability is not so intuitive sometimes, haha