I can think of 3 reasons why this can happen

-as mentioned in the catalog that some tables take quite a long time, i also noticed that in the 2 range QAction 14001 can take a while if the connection isn't great
This could cause a queue buildup of requests which overfloods the device/dataminer and therefor it couldn't set up the connection again

You can verify this with a wireshark + element log/Sniffer to know the difference in communication towards the device

A solution for this would be to make the driver less sensitive for these kind of things or to increase the timeout

-Some devices are suddenly using negative SNMP Request IDs, which is actually not supported/allowed for SNMP communication, this is a firmware issue, you can verify this as well in a wireshark that the Request IDs we send are positive numbers but we get the exact same Request ID back but as a negative value

-Another thing we have seen before is that some ports are blocked and that you by accident use a blocked port and the other external tool will automatically take another port.

I would only check the latter one if the first 2 aren't the root cause. Here you'll need to do an analysis if there is a difference between the specific port used by DataMiner and by the external tool