I think sieving 5M10M is about right, because:
1. Unlike sieving k*b^n+c where the efficiency of the sieve is greatly increased by sieving a large range at once, with the current sieves for this project there is not as much lost by sieving in stages 5M10M, 10M15M, etc.
2. A big motivation for this project is that there is no known prime of the form we seek. If we find one, then there may not be very much interest in continuing the search to find a second one, which means any extra sieving would be wasted.
3. It is quite possible for the sieve to become 50% faster in the future, but quite unlikely that LLR will become 50% faster. Because of this it would not be a good idea to spend too much time in sieving too far in advance of LLR.
