Examine the circuit reactor's congestion control-related dequeue behaviour
-
@nickm started a discussion: The trouble here is that we shouldn't always be checking the congestion window? I think that's worth having a ticket for.
The reactor code in !126 (merged) only dequeues messages to send on streams if the stream congestion window is nonzero, even if there might be messages that wouldn't count towards the congestion window. In other words, if the congestion window is zero, it's not possible to send any type of message, even those that don't care about congestion control!
I suspect this isn't that much of a problem, given the window shouldn't really be zero very often under nominal network conditions. However, this is something worth investigating in more detail.