We are investigating the cause of compute jobs failing intermittently. This is affecting all regions.
UPDATE 2018-11-19 19:45:
We've been sitting on this one for a few weeks as we have had to take several steps to alleviate the problem. We've wanted to make sure all these steps are solid and that our users are getting a solid experience before giving the green light. Happily, our last round of tests on friday passed so we believe compute jobs are now reliable.
Thank you for your patience on this complex matter.