I ran SiDock until recently. But I desire a work buffer of ~0.4 days, because this is about the period at which I can intervene if my cable connection acts up, or if the project has got issues.
SiDock still has got a limit on tasks in progress of 2 per active logical CPU, counting up to at most 64 CPUs. It is possible to work around this server-enforced limit, such that even my 88-thread and 128-thread computers can have a 0.4 days deep work buffer. (Most who read this know the details how I work around it.) The weakness of this workaround though is that a 0%-resource-share backup project no longer works as designed.
There are other Covid fighting projects, and I am now running one which
– does never run out of work during random nights,
– supports a reasonable amount of work in progress per CPU.
When the current SiDock contest is over, I'll go check how work demand and supply at SiDock develops and probably ask the admins at the SiDock message board what their plans regarding the above two bullet points are.