Regarding the non improvement to lyra2rev2, the thing you fail to understand is that code hasn't been rushed into release, but benefit from the various rounds of optimization I did on lyra2re (classics), so instead of releasing every days half broken stuff, I release once the work is over, has been tested and which I consider as stable. So beside fiddling with kernel parameter more than I did already there isn't much to do to optimise it. (new optimization are found by rethinking the code and rewriting in full or in part the logic... fiddling with kernel param is ok, but at one point you have to decide for one over an other...).
So yes you didn't see improvement. Must add when a team ask me to write an algo for their coin, I try to release something which works and is stable no some half optimized broken stuff which wouldn't be very professional...