There is one slight hiccup in the "long term it evens out" mantra, which I myself *normally* use when recommending people use PPLNS over PPS. Neutral/good luck today is worth a lot more than it will be next difficulty. Similarly, bad luck today will hurt you a lot more than it will next difficulty. So there is a potential gain/loss due to variance. Again, the odds are just as likely one way or another. It's a matter of risk assessment.
This hiccup is only present during times of continual/rapid network difficulty growth.
I agree with that, pool size needs to be large enough such that variant risk is acceptable within a 2 week time frame. A 1TH pool is very risky these days, but as you get to 50TH and up, I feel you're pretty safe. Of course that minimum-pool-hash-power-to-feel-safe number is somewhat subjective per individual miner and rises with difficulty. 6 weeks ago, I would have felt safe at 10TH and up.
As difficulty continues to rise, this becomes an issue because miners will naturally consolidate to larger pools. The reliability of the distributed network is at risk if small pools start closing down and large pools become a target as nearly a single point of failure. If the small pools aren't around as fail overs there could be major disruptions in transaction processing (and miner income) if there are any serious attacks against the network.
That's why I'm of the mind to recommend supporting smaller pools before its too late. Miners with large investments in hashing power would do well to protect the long term viability of their investments by supporting smaller pools to make sure they don't go under.
If you have a bunch of hashpower you have two choices as I see it. Reduce short term risk by mining at a huge pool and risk future network instability, or mitigate short term risk by utilizing multiple smaller pools and reduce long-term risk to the network and bitcoin in general.
