It would increase their revenue

The increase would be, for a block thar AntPool mines just after their own,

(Block total fees) * (AntPool share of total hashrate)^2 * (time to get new block and prepare new template) / (10 minutes)

And that would always be a template that is valid.

For any previous block whatsoever (not mined by AntPool) we remove the "^2". But then, it's not a guarantee that we try to include already included TX. In this scenario they should try and mine TX that are further down in the mempool queue and thus have a higher probability not to be included in the block just mined.

Anyway, for the first scenario the number is around (I'm guessing some numbers)

2 * 0.25^2 * 5 sec / 10 min = 0.002 btc = 200 ksat

Per block (all blocks even the ones they don't find)

Now that I calculated it... It's actually very significant.. why don't they do this? I guess they will start doing it now that the fees are so high. It will take some work but seems definitely worth it to get a 200k * 1/.25^2 = 0.032 btc increase in revenue from blocks that they do find