AFAIK there is a problem with fork management. As I understood, a block is marked as invalid on the side-chain (maybe relatively to the local chain), and Duniter won’t reconsider this validation when solving the fork. A workaround is to restart the node (these falty blocks are in RAM). I have a cron task that restart my node every hour whith duniter stop ; duniter start. I still have this kind of problems once every months or so, on a Pi with 2Go RAM. Before the CRON, it was every 2 days.
see this post: Désynchro rapide de mon serveur Ğ1 - #2 by elois
My node regurlarly logs lots of timeouts. It does not prevent it to work preperly.