I did read an article a while ago (can't remember where, but it was on a link from a forum discussion (probably rcgroups) on this subject) where the "official" method of long term storage for NiCads was to totally discharge them, and leave them shorted, going through a long reforming process before use. I took that to mean that different internal chemistry occurs between actively discharging a cell and just letting it discharge on the shelf. Notice "cells" rather than "batteries". In a battery, under discharge, one of the cells might be discharged before its mates, all things not being equal, and would try to reverse charge from the current produced by its mates, which would in turn prejudice against its full recharging, the probable cause of memory.
A good automatic discherger/recharger is probably a good investment, since it "should" sense the points at which the different things happen, and move to the different stages of the job at the right time.