Previous to NPM 10.7 you would poll a node via ICMP. While a node might be ping-able, that did not mean NPM was able to gather stats via SNMP. In other words, you would loose true visibility of your environment and have no idea it had happened.
In NPM 10.7, Solarwinds introduced the ability to poll a node via ICMP OR SNMP. While this is a fantastic option, we quickly realized this was still not ideal and we could still not be aware of the true state of our environment. In the case of polling a device with only SNMP, it the SNMP process had an issue, the node could appear offline, which is a lot more critical than the SNMP process daemon just having issues. But without an ICMP to go along with the SNMP poll, you wouldn't have a clue - SNMP might have failed making the whole system appear down.
So we'd like to have the option of using ICMP as the up / down polling method, but have NPM track the SNMP polling method as well and as soon as data is not longer returned - report it as a SNMP polling issue, not necessarily as a node down issue.
See thread here where I discuss it - NPM Alert on Node UP (Ping) SNMP DOWN (unknown / unresponsive)