Thx, if there are any regressions on new versions pls report them as soon as possible, it's welcome and speedups development.
Regular disconnects are bad ofc I'm not sure why you're getting them. It's most likely not an issue on the pool side, since flypool works fine, without disconnects for other users.
All GPUs are separated/independent, this design has a lot of benefits. So for example if you have different GPUs with different speeds in one system - each GPU will receive it's own job and it's own difficulty according to it's speed. Currently I'm not taking full advantage of this design, but it's planned to restarts each GPU separately on connection, hardware etc. failures.
If you have the same issues on 0.5 I'll think about a way to debug this issue, just report it.
Edit: Disconnects happen very regular, after the same amount of time, that's very strange.
I switched miners to ewbf to test, and the graph I got was no more disconnections. This has to be some sort of throttling, I have 24 cards, I split them in groups of equals. So tot he pool it must seem like I am making 24 separate connections at the exact same time.
you can see the difference between zm disconnecting, and ewbf staying normalized. Something is amiss, but I don't know exactly what it is, nothing changed, only the miner program.
https://image.prntscr.com/image/SkRIFTAmQOWRi5WTz_2ang.png