Multiple clients completing the same WU

[bikeaddict]

Here are some WUs that four clients completed in the same day. The project has a timeout of two days. This looks like something malfunctioning and reassigning the WU way before it times out. It's a waste of computing resources to do the same work multiple times.

https://apps.foldingathome.org/wu?p=182 ... =1&gen=157
https://apps.foldingathome.org/wu?p=182 ... =3&gen=195

It would be easy to see how many more there are in your database with a query of WUs with code=Ok, date(assigned) = date(returned) that groups by PRCG and having count(*) > 1

[muziqaz]

That is probably dynamic timeout feature at play.
If I tried again to argue of its removal as a default with the main fah dev, I would probably get banned

[arisu]

What is the dynamic timeout, and how does it work?

Surely setting a floor that the dynamic timeout cannot go below would help.

[muziqaz]

Dynamic timeout is a "wonderful" feature which adjust timeout automatically depending average return rate for the project, or something like that. It was enabled during COVID where it really worked. But not anymore
I asked researcher to check if that is the cause and to disable it

[toTOW]

This feature made sense when it was designed during the COVID surge. With more compute power available than work, it allowed FAH to get faster returns by shortening the wait before reassigning work.

But it's completely useless now we have more work available than compute power to complete it ...

[arisu]

This feature made sense when it was designed during the COVID surge. With more compute power available than work, it allowed FAH to get faster returns by shortening the wait before reassigning work.

But it's completely useless now we have more work available than compute power to complete it ...

I thought it was the other way around and there were shortages. Or is that only for GPUs?

[muziqaz]

Recent shortages were a blips. There is no shortage anymore