Critical Description
With oak-core 1.0.8.r1644758 we're seeing revision garbage piling up - even though revisionGC runs regularly (every night) and does not complain about any failures.
This happens with often-changed nodes, ie with nodes that get more changes that fit into a mongo document, thus get split between revisionGC-runs.
As a negative side-effect of this, the observation event generating (ie diffImpl) seems to have problems with such revision garbage and performs tremendously slow - presumably because all garbage-revisions are read even though the diff should be very small (ie the fromRev/toRev are only a few seconds, if ever minutes apart) - but I'll file a separate ticket about this.