I'm using sr2sieve to sieve Proth with large k (far out of PG's range) and 1E5 <= n<= 4E6.
It seemed to me like that because I never saw that a factor turned up more than once. Now that I write it, it sounds a bit stupid... So I either just missed it or it happens rarely, but each prime is tested against all candidates?
I guess it just happens very rarely, just now I found about 400 composites with 4E12 prime factors. that means only one in 1E10 primes was actually a factor. So the chance that one prime factors more than one candidate is very low in that range.
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 4,400,000