I have several masternodes which are in "idle mode" since the customer did not start them yet. It seems these idling nodes have a high probability to get stuck in ~6h. Active masternodes are doing fine though.

I reviewed the log of some of the stuck nodes and they all have the same pattern. Below are the last lines in the log:

node01

[...]
2015-08-21 02:45:41 CActiveMasternode::ManageStatus() - Checking inbound connection to ... 

node02

[...]
2015-08-21 03:22:48 CActiveMasternode::ManageStatus() - Checking inbound connection to ...

node03

[...]
2015-08-20 23:03:53 CActiveMasternode::ManageStatus() - Checking inbound connection to ...

So all nodes seem to be stuck at the same routine