Statistically speaking, the time to mint an x-bit stamp is highly
unpredictable. On average, you can mint an x-bit every 2^x/s seconds
(where s = hashes/second), but the actual duration commonly ranges
from barely any time at all, to 2-3 times the average mint time.
Minting *eight* stamps of postage x-3 would correspond to the same
average amount of work, but, statistically speaking, would be *much*
more consistent. Having to wait unpredictable times for stamps to be
generated is annoying.
It's also only eight times the data transfer cost, so not a big deal.
Thoughts?