While none of the management operations _should_ affect normal operation, there have been numerous bugs over the years in which SNMP polling as well as interactive command operations have caused CPUHOGs, crashes, performance problems, etc.
HUM only uses SNMP to gather its data. As long as you don't overwhelm devices with polling traffic, you shouldn't see problems, but I can't say for certain that you won't see problems.
If possible, you should test your pollers and templates on non-production devices.