Orphaned blocks occurs when two blocks are sent simultaneously to nodes and as miners mined or build blocks the longest chain one is been picked by the nodes and the later dropped back to the mempool. So if your transaction happens to be in the short chain then it can be reversed.
The correct term here is stale blocks. Orphaned blocks refer to blocks without a known parent (as the word "orphan" implies), which is something different.
Also, a transaction which is included in the losing chain will only go back to unconfirmed if it is in neither of the block at the same height or the block at height +1 in the winning chain. Otherwise it will remain confirmed despite the chain split.
The sender can increase the transaction fees and also change the recipient of the funds which will make the former transaction invalid (Replace by Fee (RBF).
No, it won't. Broadcasting an RBF replacement does not make the original transaction invalid whatsoever. The original transaction remains entirely valid and could still be included in a block if a miner chose to do so or did not learn about the higher fee paying replacement.
if a pool only matches the speed(50%) of the rest of the network it will never catch up and always be 6 blocks behind
Your numbers are all wrong.
If you take the equations from section 11 of the whitepaper and plug in the numbers for an attacker with a probability of 0.5 of finding the next block (i.e. 50% of the hashrate), then the probability they will catch up from any deficit is 1.
You are assuming that both attacker and honest miners will mine blocks at a regular 10 minute interval. Given that there is random variance, and given an unlimited amount of time, then the attacker will in fact
always catch up. With honest nodes simply following the chain with the most work, then honest miners have to stay ahead
at all times, whereas the attacker only has to take the lead
once for all honest nodes to then switch to their malicious chain.
Further, a 60% attack does not mean the attacker is 10% faster. If the attacker has 60% of the hashrate against an honest 40%, then the attacker is 50% faster than the rest of the network.
A 70% attack results in the attacker being 70/30 = 2.33 times faster, and an 80% attack results in the attacker being 80/20 = 4 times faster.
as you can see even after doing over 20 blocks costing millions of mining power you may only get to overtake within 20 blocks if you had like 80% of hashpower
Again, not true. With 51% of the hashrate you are mathematically guaranteed to overtake any deficit. You certainly don't need 80%.