The problem could be the heat sinks but it seems that bitmain acknowledged this problem with the s17 series and supposedly fixed the issue on the S19.
I do agree, after a second thought, I just remembered that the S19 have large heatsinks that cover more than one chip and it's tightened with some screws, so it's unlikely. but still, one bad chip can cause a mess in these hash boards, so I would still do the board by board troubleshooting if they miner still acts weird even after you move it to a new location.
Here is another miner. It just now stopped mining. Kernel Log is similar to all the other miners with the same problem.
2021-01-19 23:52:22 freq = 525, percent = 90, hcn = 12480, timeout = 449
2021-01-19 23:52:22 set_start_time_point total_tv_start_sys=270 total_tv_end_sys=271
2021-01-19 23:52:22 set_voltage_by_steps to 1292.
2021-01-19 23:53:47 set_voltage_by_steps to 1272.
2021-01-20 20:18:42 set_voltage_by_steps to 1260.
2021-01-20 21:45:40 set_voltage_by_steps to 1282.
2021-01-20 21:53:27 avg rate is lower than ideal rate, 72720.92 in 30 mins
2021-01-20 21:53:27 chain 0 hash rate 11401.00 low in 15 mins
2021-01-20 21:53:27 chain 1 hash rate 10699.00 low in 15 mins
2021-01-20 21:53:27 chain 2 hash rate 11628.00 low in 15 mins
2021-01-20 22:08:28 chain 0 hash rate 0.00 low in 15 mins
2021-01-20 22:08:28 chain 1 hash rate 0.00 low in 15 mins
2021-01-20 22:08:28 chain 2 hash rate 0.00 low in 15 mins
2021-01-20 22:23:29 avg rate is lower than ideal rate, 0.00 in 30 mins
You can see that the hash rate slowly came down to zero with no reason. No errors, temps are good. Worked fine for 22 hours. Only happens when outside temps go to 50f degrees. Any other tips from anyone please leave them. I will inspect the copper wires for any loose connection in the outlet and the breaker.