For simplicity, we assume that the only value in the block reward is the mining fees.
For simplicity, we assume that there are no limits in block size, and that a reversing miner can take all fees in both blocks.

E(mining block) = sum of all fees on current block * p(mining block t)
E(mining previous block to realloc fees) = sum of all fees on current block * p(mining block t-1) * p(mining block t)
E(mining prev) > E(mining current) when:  prev fees * p(mining) > current fees.

I'm pretty sure that p(mining) is equivalent to hash power, but I'm tired and I might be forgetting something.
Assuming no individual party has more than 50% hashrate, E(mining prev) will never be greater than E(mining current) so long as the fee payout in the current block is at least half of the fee payout in the previous block.

So, if all transactions contribute to a 'fee pool', and the fee pool pays out exactly 50% of its contents every block, then this type of attack never makes sense, because if you had 50% hashing power you could perform more interesting attacks.