I definitely agree with this in principle, and have found it to be very
effective in practice. However, I've been using SRS software for about
eighteen months now. At the six month mark, I reset my database and
started over. One reason is that I switched from SuperMemo to
Mnemosyne, but another reason is that I had decided to make my cards in
a different (and I hoped at the time, more effective) way. At the same
time, I kept my old cards, though I started them over again from 0.
Although I began studying the same material again, this switch really
helped me, and I certainly would not have done as well had I kept my old
rep data.
So, although I definitely agree that the best practice is to trust the
algorithm, there are some exceptions.
Regards,
Patrick