We’ve been seeing this on and off with our K2 Five processes at the moment. For a while it baffled us as to what was causing it and it seemed to be pretty random and retrying the step resolves the error.
After a bit of digging we found that it was related to a race condition generally in our development environments where we were responding to workflow tasks so quickly the server hadn’t had time to finish setting up the task in the background. This is not a normal situation and in a production environment the user isn’t sitting there waiting to immediately jump on a new task and action it. If we gave the server a normal production type experience that a normal user would give it, the problem didn’t occur.
Hopefully this is just one of those weird development situation. We’re keeping an eye on our prod environments to see if it happens there.