In response to a change in the brain's biophysicochemical stew (meds, trauma, chronic stress, etc), it generally takes neurons a couple weeks to fully generate new or recycle old protein machinery -- to adapt to changes. This duration can be shorter for some proteins, longer for others. Thus, small changes would be expected to minimize the shock to the system. This just makes good homeostatic sense.
There are problems with this as dogma, however. There is not useful research, at least that I am currently aware of, to demonstrate whether the "best" interval percentage change is 10% or 5% or 25%. We also don't know if the "best" interval is one week, two, four or eight. Or which medications and their affected pathways are best tapered at what intervals and amounts. Please share original source (ie, PubMed) links to peer-reviewed research below if you have relevant references.