You note that when using a MID server cluster that you'll receive timeouts during SNMP Classify on various IPs. This issue may also be observed to occur less frequently as the number of MIDs in the cluster decreases and does not occur at all when using a single MID server.
This issue is primarily observed when running SNMP discovery against low resource targets such as UPS devices
When setting mid.log.level to debug and checking MID agent logs you see the following logging around the timestamp/request that times out:
SNMP Response was not received within 400 milliseconds... Resending request
Release or Environment
The default repeat_interval to resend an SNMP request when we don't receive a response is 400ms. We then repeat the request until we reach the timeout interval.
For low resource devices they do not respond all that quickly and then combined with the processing time it takes for the platform to determine cluster routing logic we've reached the 400ms and re-send the request which repeats the behavior ad nauseam until the timeout is reached
Create/increase the "repeat_interval" probe parameter on the SNMP Classify probe record until you're no longer consistently experiencing timeouts. If you don't want the probe request to repeat at all you can set repeat_interval to match the configured SNMP timeout value.