Basically, as many as you have resources to protect. The server/node architecture of Sandfly can be distributed using named queues and jump hosts to scale as needed. Each node for instance can protect many thousands of hosts with minimal  hardware resources, and new nodes can be added to scale upwards.

Servers are limited only by the CPU/RAM you wish to dedicate to load scaling. Multiple servers can be used to distribute larger deployments. You can also choose how much and where to store forensic data according to your requirements. For context, we have customers using Sandfly to protect tens of thousands of Linux hosts.