Seems to check out in reality as well: my dedicated miniPC that usually takes 15-16 hours to crunch a v0.05 task is on track to finish these new ones in exactly 30 minutes.

I think the explanation is that the server shows WUs as pending when both tasks have been reported and the validator is overloaded. In your case the wingman's tasks are still in progress thus the WU hasn't even reached the validator.

The whole work process seems to be very cyclical, each cycle taking roughly a week:
1. Large batches of work gets generated
2. Many rigs jump to crunch whatever they can grab
3. Tens of thousands of WUs get crunched in 1-2 days
4. Validator is overloaded once crunching nears its peak
5. Server runs out of disk space
6. Task upload and reporting barely works for 1-2 days
7. Work generation is stopped
8. Validator catches up in 1-2 days (including tasks that get re-sent because of missed deadlines)
9. All WUs get crunched and validated, project runs out of work.
10. Repeat from 1.

I'm thinking the server is slowly freeing up space after validating/deleting and then it allows a few more tasks to be uploaded/reported. Since there are hundreds of rigs waiting to upload and report, it takes time for everyone to get through the bottleneck. Keep requesting updates from time to time until you find an opening.

I believe DENIS-fiber could be highly efficient on GPUs. If there's interest from the admins, I could help with porting it.

This could be another big help from you. Thank you in advance!!

I believe DENIS-fiber could be highly efficient on GPUs. If there's interest from the admins, I could help with porting it.

I know, but since you assumed in your initial post the points are related to runtime, they should be different between linux and windows with the old app version, right?

So I would assume there is something different which results in lower points per wu.

I’ve not changed anything from the simulations code, only for the compilation.
Hope this helps!

If the work needs doing I do not consider it a waste. It all is part of testing things

Good morning!
Finally I have been able to move the compilation of my local machine to Docker, it has cost me because I had problems with the dependencies that were really necessary for the compilation of BOINC, as it is constantly changing the requirements that are asked to compile, the previous simulator was almost obsolete, and as locally with the visual interface is always installed more, I wanted to optimize as much as possible the compilation.

I just kinda gave up on it. I am not going to baby the app clicking to request work over and over again.