Heartbeat is not yet GA. Since 5.3 there have been a few changes to the event structure, but I don't think any changes that affect "capacity". Network throughput for the machine and the maximum number of open FDs are probably the limiting factors.
I would check the ulimit values for the process and also set the
schedular.limit option appropriately.
For GA we would like to have an I/O rate limit config option which could help when monitoring lots of services. And we might also want to add some DNS caching as that can also get expensive when monitoring lots of services.
I recommend giving the latest release a try and seeing what issues you encounter, and we can make sure these are accounted for in the roadmap.