There is a lot of discussion of the propagation of publishing blocks and reacting to the publication of a competing block. In this strategy, the 'selfish' pool could take a probabilistic approach, and simply hold on to their private block for "a while" based on the probability of someone else providing the competing block (obviously still publishing instantly if a competing block is found).

This means they still get some advantage, in a time lead over the general network, but they balance out some of the costs and risks of waiting. They will get their lucky block-ahead of the network less often, but perhaps it is a better strategy. Fascinating stuff.