The first (and smallest) issue has to do with my Thinlinc installations. The agent component can hang after running a specific cronjob. A "ps" process starts to hang. The only way to resolv this is to restart the zone. This behaviour seems to be related to the lxcfs package on version 0.13-pve3. Later versions seem to resolve this.
The second issue was much more of a problem. This manifests itself as a server with very high load (>12), lots of processes (normally 1400, now > 10.000) but otherwise responding fine. Everything that is running in the lxc containers seems to hang on IO. While checking the processlist further (before rebooting the server, as this seems the only way out), the output of pstree shows something intresting.
So there seem to be 671 threads running for lxcfs? Normally this is around 10 on this system.
After downgrading the lxcfs 2.0.0-pve1 package to 0.13-pve3, this issue seems to be gone. Downgrading is very straight forward:
After the downgrade, it seems best to restart the server, so all containers run with this version of lxcfs.
There now is a new version of the lxcfs package, 2.0.0-pve2. It looks like this package fixes the above two issue. I've currently not seen any new issue's :)