I have a question, say I broadcast a transaction with an RBF flag, and then try to replace it with a new version at a higher fee.
Is there not a case in which some nodes that don't support RBF reject it as they see it as double spent.
I think this is the case, but you never know until a service (node operator) tells you so. During one of the recent spam attacks (when high priority fees were reaching 1000 satoshis/byte), I submitted a stuck transaction to an acceleration service on the forum. It was an RBF transaction that had already been submitted with a higher fee, since the network fee rate kept increasing.
I was told that the transaction was a double spend, so it couldn't be accelerated. I'm curious which pool the service was submitting to, but I can't find the original post now.
Now you have some nodes that still contain the original transaction ignoring the new one and is there not a chance that it could still get minded?
It's possible that either transaction could be confirmed, depending which pools mine subsequent blocks. I've seen two transactions spending the same inputs, where the lower fee transaction got confirmed (and the higher fee transaction got discarded by the network). Pools have transaction priority mechanisms that we'll never be privy to, and users on the network can always submit the lower fee RBF transaction to an accelerator service. In that case, fees aren't everything.