After improving and fixing digests memory usage and robustness, it seems that cron execution is stopped 3-4 times each day. After some debugging, it seems that some users cause the process to die every-day (note that it isn't critical because, after
MDL-11657, things will continue in next execution).
But it's far from perfect so I'm debugging a bit more cron execution in order to detect what happens with those users.
Some ids causing problems: