I would like to know how he/she did it !
For every 5 bits range increase about 17% of previous range covers it, but to reuse DPs means using same jumps, so more jumps on average per kangaroo = more total operations than optimal.
If you start searching in 135 bits, then it will take the same amount of time to find private key 0x1 or private key 2**134 - 1 or private key 2**89. So if we adapt the jumps needed to go through 135 bits and solve a lower interval than that (as if we would try to jump through 135), that means longer jumps on average per kangaroo = more total operations needed than optimal.
I think it's pretty obvious what the 120 / 125 / 130 solver has done. You won't find the software publicly, you should all forget about that ever happening. Learn to code, this is what this competition is about first and foremost, not the prize.
It depends on how you plan to reuse them.
You are looking at it from one perspective, to run the original pub, in its original range, with the DPs generated from a lower range.
There are 9 million ways to skin a cat.
I am sure you have done this kind of test and analysis, so answer me this, if you reuse DPs found during a 66 or 65 or 70 bit range, to find a key in the same exact range, how much search time did it take, was it less, if so, how much less, on average?