Been thinking about this one for a while as I've noticed the only coordination is for large n (>9M) where we are P-1 testing just before PRP gets there.

How about a repository of k,n pairs with known tested B1/B2 bounds and maybe even the save file from Prime95/sbfactor/etc.

I've been playing around with this, seeing how much storage would be required to store save files for each outstanding k,n pair. It's feasible, although I'll have to jump through a few hoops to provide simple access for other people.

This would save much duplicated effort.

Just thinking about this at the moment, comments welcome.