When starting a simple 3 node cluster, after about 5-10 minutes, the "leader" dies throwing this error:

Accept error: accept tcp [::]:4001: too many open files

Using debug output, the only thing it shows is a sudden stop when doing:

[raft]01:28:53.821293 peer.heartbeat.flush

I'm not sure where to look. I'm experiencing the same issue in a separate Raft based project and wanted to try the reference implementation which appears to be doing the same thing.

Any ideas?