we are not subtracting the target we are subtracting from the target. Ps(Strat range) Pt(traget) pb(point of scalar 1billion).
PS +Pb =Q , Pt + (-Q)=R , then we look for x of R in Db .
What do you mean by "the target point needs to get subtracted in sqrt(rangeSize) delta increments".
I know this strategy can't help much with 135bit , that is why I have another approach with that Db
It's faster to simply do Pt - j*Pb = R and look for x in the precomputed [startRange, startRange + 1 billion] table.
This is basically how BSGS works.
If the private key is beyond 58 bits this gets exponentially more inefficient, the larger the range is.
For 135 bits you'd need a 2**67 database and 2**67 subtraction & lookup steps. But due to space and time constraints this is not possible for the foreseeable future (e.g. our lifetime).