Monitor physical network gear and cloud infrastructure side by side, without treating them as two separate tools.
Your on-prem collector polls routers, switches and firewalls over SNMP, SSH and syslog exactly as it would in a fully on-prem network.
Add cloud VPCs, load balancers and managed databases as REST API integrations rather than trying to route SNMP into a cloud VPC.
A group like "production" that spans both on-prem and cloud resources is often more useful for alerting than splitting strictly by where something physically runs.
The connections between on-prem and cloud, VPN tunnels, Direct Connect or ExpressRoute links, are common blind spots; make sure they're monitored as devices in their own right, not assumed to just work.