Rudder

There a bug in the current "save node configuration cache" method that delete node cache for not modified cache. So, in some case, you could hit cycle where a part of nodes are saved (removing the other part), then the other part is saved (removing the first), etc.

Of course, all forseable bad consequences happens: increased generation time, patological config id modification without actual changes in the expected reports (and yes, we get bad results with compliance), etc.