I like the approach taken by Ian Jobson here
I try to do what he does (divide colonies into 3 categories) because it seems like a way to make headway for somebody like me with a relatively small number of hives. Breed from the queens in the best third & re-queen those in the bottom third, and remove their drone cells.
Additionally I try to re-queen colonies where they swarmed this season with a 1 year old queen (or would have swarmed without my intervention). I don't know if I'm being too harsh on the swarmy ones - what do others think?
When viewed as populations rather than individual colonies this is a strong method of pushing desirable characteristics forwards without losing diversity and vigour.
If you picture a population's characteristics as a bell curve with poor on the left and good on the right of the bell, then obviously the aim is to shift the meat of the bell right. The easiest way to do this is to cull the unproductive tail and propagate from the best third.
I try not to use the same breeder queen for more than two grafts a season( a total of about 80 daughters with my system) to avoid inbreeding suppression. It's the following seasons drones that's the issue, if all the drone mothers originate from the same mother then I can envisage losing sex alleles from a population even with open mating.