These tend to fail quite often on Travis due to multiple reasons. One
reason is that establishing intra quorum connections take some time and
the tests in dip3-deterministicmns.py did not sleep long enough. Another
reason is that the individual stages were not really checked for completion
but instead just a hardcoded sleep was used. And another reason was that
with a total of 13 MNs, it's not guaranteed that every DKG results in one
MN to be punished.
