Hello Mr. M
Software is fast and smooth. But it will consume large amounts of network, resulting in network congestion, can not open a web page in normally.
I tested 50 ants S3, use 64 threads, display refresh one second refresh cycle 300 seconds. 1 second to 50 units will be able to check the working condition of ants, everything is normal..
If I change the refresh cycle is 100 seconds, you will find there are 4-8 sets of ants S3 displays an error.
Try two threads, still error occurs. I do not know it is the core network segment VLAN switch problems, or other reasons.
Test 1,000 ants S3, first pause about 5-10 seconds, check to 30 mining machine working status after an error occurs. Finally, a large area of error and  not working. Network is not down, you can not browse the Web, other computer turns ants console page.
I think it is a program to check the dozens of mining machine, got a job status information, but these threads still remain and mining machines to connect and not disconnected, causing network congestion.
You can try your method, do a version
I also have a stupid way like this
Option 1, for example, 64 threads, according to the IP address has been added to check the order of 64 mining machine working condition, after get a result, if there is no machine need to restart, then end off this 60 threads, pause for a few seconds, in order to continue check mining machine operation, until checked all mining machine.
Option 2: You can set the time to detect 50 mining machine (IP2-51, 300 seconds after the break, continue to check the working status of 50 mining machine (IP 52-102), loop executes until the examination was part mining machine.
Many thanks for your hard work, you've done well, I'm here got many  machines, multi-network segment. You do not have the embarrassment, you have a good technique. You and your M monitoring are great!