Running 0.4.4 for about 5 hours now, so far so good.
Is there a reason the GPUs slowly change order in the output, can they be fixed in place instead so always output in order 0,1,2,3.... etc.?
The GPU that finishes first prints it's output first, they are all independent. So you're able to see how things actually behave on your system. This can be changed if it's important.