Running Sentinel 7.2

I have a setup that looks like this that collects logs from a debian
wheezy running rsyslogd:

Custom collector (parse a specific application)
->Syslog connector tcp
-> event source based on appid

Universal collector (parse general unix events)
->Syslog connector tcp (B)
-> event source universal (A)

I'm trying to configure a cache on the debian server in case the
Sentinel server becomes unavailable and during my testing I've seen some
odd behaviour that I don't know if it's by design or not.

If I stop my event source (A) and then start it again I see no events
coming through A or B when looking at the raw data.

I'm running tcpdump on both the debian and the sentinel server and I can
see packets being sent and received but nothing is going into sentinel.

To get the sentinel to recognize the data coming in I have to restart
the syslog connector tcp (B).

After restarting the syslog connector tcp (B) everything works again.

Is this by design? I.e. should I avoid stopping event sources?