This Dry Run is now over... here is where we are at:
It appears that we resolved many issues and that the primary cause of forks right now is network propagation delay caused by CPU load, bandwidth throttling, and the single-threaded nature of the code.
I would like to separate logic bugs from performance induced bugs because we have ways to significantly enhance performance over time. Dry run 6 will focus on eliminating the effects of performance from the equation while introducing many small enhancements at the same time.
Stay tuned.