I'm having severe stability issues with this miner on my rx580 4gb cards
Some algos select an intensity that is too high by default that uses more than 4gb of vram and causes the system to stall and reboot (lightv7 for example, defaults to 130? but 110 is the maxium, but on normalv7 it defaults to 55 and can be increased to 58 without issues).
But even after tweaking all config files, I can't never get a uptime of 6 hours, it always reboots without no log. I've reduced the ram from 2100 to 1960mhz, increased vcore, nothing helps. It just restarts the machine.
Sometimes I notice that 1 thread is randomly not running. I have 5 GPUs and you see 330 620 620 620 620 H/s on the logs (random example).
I used to have 1 month uptime with ETH mining (rebooted just to install security updates). Running on 10 LTSB with 18.6.1 drivers. I'm trying to debug this as best as I can because the profit is near 0€ per day, every second of downtime is bad.

Suggestion: the configuration needs improvements because multiple files are a mess (if you need an API to monitor, you have to write that on every config file for every algo), default intensity needs tweaking, gpu watchdog could be reworked to search for failed threads instead of a failed gpu, on pools with multiple algos, the miner should first check the pool for the current algo being mined and then load the kernels (right now it loads normalv7 kernels, connects to pool, loads "correct" kernels).