If you need to find a temporary workaround, I think promoting the
server
to be a supplier could be a workaround. Drawback is that it will store
changes in changelog but if there not a lot of changes it has no disk
impact neither significant response time impact.
I have enabled replication on the instance with the role 'master'
('supplier' was not available) :
sudo dsconf <INSTANCE> replication enable --suffix "<SUFFIX>" --role
"master" --replica-id 1
and then restarted everything.
But it still crashes as soon as compaction is triggered.