In article <5cj421pp4js7ueaj2a0htbjq3800hvgvj7@4ax.com>,
ffgeorge@yourplace.com wrote:

In Classic a unit could have been sent out a few hundred times and no

Typically it was about 3 or 4 times, but maybe more if there was a shortage
of work units from the splitters at the time.

In its final form, classic S@H sent an initial burst, then resent when the
work unit became the least recently sent, purging it from the candidates
when it was verified (see below) and space was needed.  The main difference
in SETI/BOINC is that the client is stalled if there are no unverified
work units available, whereas the classic version re-sent work units to
keep the client busy.

one new if it was crunched properly or not. Those who overclock and

Berkeley knew if work units were correctly processed (in the sense
that multiple people returned the same results within the limits of
the calculation accuracy - i.e. the same test as weith BOINC/SETI).
Credit was always initially allocated before cross-checking and, at
least originally, was not revoked if the results failed to validate.
This had an advantage in discouraging unofficial optimisations as it
meant that one couldn't tell whether the results would eventually be
counted based on the intitial score updates.