You're conflating a lot of things here. Racing a fraudulent transaction to have a legitimate one confirm faster is not a double spend, like at all. Up until the moment one of the transactions confirms, no funds have changed hands. What you're also conflating is the fact that the hash collision has no influence on transacting funds in the first place. The DCI report recently released their vulnerability exploiting code which demonstrated, that in order to actually use colliding transaction hashes to create fraudulent transactions, the fraudulent party would have to have access to the seed at which point the whole thing becomes moot anyway. Adding to all this, the hashing function has long since been replaced, further compounding the moot-icity.