After a very long support case with jtac i was told there is a bug in the version i am running.
Greetings from Juniper.
We have received an update from our engineering team that as per the data collected during the issue , they can see that the session-sync data is not synchronized between the nodes and the counters on node 1 was continuously incrementing and this causes the packet drops on node 1. We already have an old PR to track this issue and this is fixed in following versions 12.1X44-D30 / 12.1X45-D25 / 11.4R11 / 12.1X46-D10. We suggest you to upgrade to one of these versions .
Feel free to contact us incase of any questions.
I have not yet upgraded so I can not cofirm it fixed the issue.
exactly the same problem, but without cluster and for source limit for session, junos 12.1x44.D11.4. It was normal, that IPS/UTM and additional functionality like nat multipoint tunnels are always with bugs, but it's too bad that even base functions like name resolving or screens are broken in new junos releases.