I know the spotty up-time over February was a bit shit, so I've come to share some details. We've had to change the way the .net site operates and service seems to be solid now. The infrastructure endchan uses didn't change, but some of the underlying hardware did.
From the client perspective, you would have received a TCP reset packet which manifested in your browser as a "connection refused". On the server side, this was caused by intermittent failures with iptables. Semi-randomly traffic wouldn't get redirected and the system responded with a refusal as result. This behavior (or lack of) could be temporarily resolved be restarting the networking stack entirely.
We've reworked the pipeline to remove the dependency on iptables all-together. It's not a true resolution, we believe the root of the issue may be within the implementation of the network stack provided by the virtualization platform.
Sorry about the interruptions, things should be stable now but if you're experiencing issues please let us know.