Meni, thank you very much for your analysis. I can't judge if you are right, but I also have the feeling, somethings wrong with the MaxPPS-method. I think it would work, if old shares wouldn't count forever. If for example all shares from more than 20 rounds ago wouldn't be considered anymore such problems as you described should not arise.

Anyway, if Eclipsemc really uses your scoring method [1], my problem will be solved and you'll find me there. ;-)
Of course I'm still sad that Continuum goes away...



[1] I'm going to find out / ask and report back.