Restarted backend not picked up (no replies)

Hi,

I am running haproxy 1.5.4 with numproc=8 (because a single process is
exhausting one core). For a service with 6 back-ends, 4 of the backend
services were restarted but but never picked up by haproxy - they were
still considered down, even I could clearly reach them from the haproxy
node with ping, telnet, etc. After restarting haproxy, those backends were
recognized as up again.

I looked briefly at the 1.5.x release notes after 1.5.4, and didn't see
anything obviously related. Has anyone had the same problem?

Thanks!
-Simon