Assert while the Nodes are down about hiredis-cluster HOT 2 CLOSED

After first having run a lots of test with node restarts I could only find a segfault issue that is to be fixed with #107, but the assert() described above was never hit.
In my tests redisClusterAsyncCommand is either called by a timer event or via an event_base_once.

Normally a command callback is called by the underlaying hiredis and when the connection is disconnected hiredis will reset the async context pointer in a node object, i.e the assert should not be reached/triggered.

But, I think I managed to reproduce the issue when calling redisClusterAsyncCommand from a command callback, i.e. when a given callback is called after a command has been sent. This callback gives a NULL reply when a connection is broken, but if it also attempts to send a new command before the callback has finished the node object still has an async context with errors.
Is this similar to your setup? I.e do you use libevent and send commands from a command callback?

from hiredis-cluster.

Thank you.

We are invoking redisClusterAsyncCommand() calls both from the application, and also from hiredis callbacks.
But we have also added a guard -- we return immediately without processing any Redis callbacks that get called after the disconnect callback indicating that the connection has broken. So the assert must be getting triggered from redisClusterAsyncCommand() called by the application.

from hiredis-cluster.