extract correlation between consumer_count and limiter registration

...and fix a bug in the process: during consumer registration we were
using the *old* cr, with the old (and usually undefined) limiter, to
check whether the channel is blocked. Thus we would end up running
through the message queue unnecessarily. No big deal, but certainly
not what was intended.

Also, make the update_ch_record return the record, which makes the
function more composable.

0 files changed, 0 insertions, 0 deletions