I don't know what is the best option anymore. Looks like slicing or dynamic penalties like Ryouiki's fork is the most efficient according to Ryouiki's simulation. I am sure you know more about this than I do. I am grateful for all of your hard work.
Strictly from a math perspective, the only best option is to mine at the pool that currently has the fewest shares in the round (ignoring score-based pools for the moment). The only adjustments to that are things like fees, donations, and perhaps rejected share ratios. (Especially when mining at smaller pools, it may take several weeks or months for things to reach a meaningful average, so short-term data is next to meaningless.)
If someone has simulations or other data showing that some other strategy is better, I'd be interested to see it. I have no access to hoppersden, so I can't really look at anything there.