May be better consider following upstream and this commit - zcash@31d49b0 , which is part of zcash#5271 ?

p.s. As the original PR touches the wallet, my offer is to postpone these changes till "after the next hardfork", to test everything properly, etc.

I think we could pull zcash code after the upcoming hardfork but merge this fix for now:
it is very simple and for new notaries it will improve node setup performance (as they will do importprivkey many times and with this fix performance increases by as much as from one to two orders of magnitude)

I think we could pull zcash code after the upcoming hardfork but merge this fix for now: it is very simple and for new notaries it will improve node setup performance (as they will do importprivkey many times and with this fix performance increases by as much as from one to two orders of magnitude)

I guess cherry-picking just zcash@31d49b0 then, will give the even more perfomance improvement, bcz std::multimap<int64_t, TxPair > will not recreate / compute on every need, but store in memory. My point of view - we don't need to add half-fixes / partial-solutions, etc. into codebase.

Also, as a notary node operator I'll recommend everybody use importprivkey without rescan, to achieve maximum perfomance and use existing set of scripts (sendfromaddress.sh, etc.) to access the funds "before" import. It will be much faster than any wallet rescan anyway.