The benefit of a shorter deadline in my opinion has nothing to do with people who actually do the work. It is for those systems that take thousands of work units and never touch them. It helps the project and the wingmen to have those units expire sooner so they can be given to a machine that will process them.

But then the task still must expire before it gets resent to a 3rd host. IMHO, a better way to handle the problem with hosts caching excessive numbers of tasks is to use the server-side setting that limits the number of tasks a host can have in progress. The admins can use the option to restrict hosts to having only 1 task in progress per core even if the owner has specified a 10 day cache. 1 task per core is probably too low, 2 or 4 is probably more realistic.

I think a 7 day deadline is reasonable for work that takes 3 hours to complete. Crunchers can easily handle a weekend of downtime on either end with no ill affect. It seems to me to be a win/win suggestion.

I agree, a 7 day deadline for approximately 3 hour tasks is reasonable. However, I think that if we request change then we should present realistic and factual arguments to support our request. IMHO, some of the arguments presented in this thread are not based on fact.

I have lowered deadline to 10.5 days.