> Choosing to publish zero transactions is not rational behaviour
Currently happens multiple times per day. Check the blockchain an empty block is published every once in a while (Although they do this because they have poor internet)
> because a miner doing so loses out on collected fees to the next miner, the same is true of cherry picking individual transactions; even if one miner does this, the next guy has no reason to do so.
Except the cost is low. My $0.04 tx fee as a single user cannot compete with various interests of censorship. However, if my $0.04 was pooled together with other users who didnt want me to be censored it would balloon into a larger amount that would be very hard to convince all miners not to process.
For censorship to be a real concern, such a transaction would never make it into the blockchain - just because a minority of hashing power decides to censor your transaction doesn't mean the entire network will; like I say, overall this is not rational behaviour for miners.