f2pool does indeed mine some non-standard transactions, but this does not lead to them 'comrpomising' any addresses or outputs. Non-standard transactions are those that bitcoin core rejects for being spammy, or for using restricted opcodes that are deemed a 'security risk'. (I never understood this logic).
Some of the Bitcoin opcodes were genuinely dangerous and were permanently disabled to prevent the network imploding. The way they were implemented meant you could have made scripts that are impossible for anybody to verify, and any node that tried would have crashed out due to not having yottabytes of memory. There is no possibility of them coming back, but new opcodes that have the same function could be created in the future if really necessary.
The scripting standardness tests are from pretty early on, but they are still around today because ultimately the scripting language isn't as useful as anybody would have liked. You can make annoying scripts and scripts that break reimplementations in spectacular ways, but very little in the way of useful ones outside of the standard set. A majority of the opcodes have never once been used in any script in the 5+ year history of Bitcoin transactions, and nobody seems to be campaigning for them to be freely relayed again.
It isn't about liberty, it is about following the protocol and one day empty blocks could being disallowed by consensus. If one of the biggest miners decides to mine only empty blocks, it could delay transactions and even when it is clear miners don't want to mishandle the network, it could happen.
That will not, and can not ever become a consensus rule. If you try to make 0 transaction blocks invalid, the miner will just pack in junk to make up the limit. It is also completely plausible that there will be times of zero transaction volume which you need to contend with.
