In article <EMn_b.162337$cM1.31158425@twister.nyc.rr.com>,
sweet <sweet430@hotmail.com> wrote:

They accept wu's which only took less than a week. If the machine takes more
than that the same wu is handed out again. Maybe the policy has changed
since I last read it.

This is a mis-interpretation of an obsolete oversimplification.

They actually send out two or three copies in quick succession.  They
then wait for at least two returned work units to agree with each other.
In the mean time, if the work unit becomes the oldest work unit in the 
pool of work units, it will be resent.  When they were last sending them
one at a time, this typically took less than a day.  Once they have 
a corroborated result, the work unit becomes eligible for deletion from
the pool, but will only be deleted if space is needed for more.

One copy of the results is retained as the official set of results, and
the identity of the corroborating results' sender is also retained.
Subsequent returns for the work unit are checked against the reference
result and credit given if they agree.

Given average processing times are probably under a day, anything much
over a day to turn round the work unit means that you will get credit,
but your processing will not contribute anything new to the science. 
However, if everyone receiving a workunit is equally slow to process
it, you could still be the primary detector after an arbitrary long
period.  Obviously the chances of this drop off rapidly with the number
of copies sent out.

If there really are no new work units going into the pool, no returns
will do any good for the science - they probably do it to keep the
work force motivated.