The new version with both tricks >>25 added works and is considerably faster than >>18. I just need to write up some sort of description. As a preview, here is R(10^20):
k 20 index 100000000000000000000
levels: 5
100000000000000000000 -> 5000000000942800098290022420982686040347 0xEB194F8ED94AC565124861C3B7D0C411B 132