Regarding the network problems we had last week, there is unfortunately not yet a resolution found.
As stated in the other status reports, we have already replaced the entire hardware (including all cables and modules). The problem recurs (though somewhat less). The assumption is currently being confirmed that some transit traffic is partially being processed via the control plane for unknown reasons. This inevitably leads to load problems, which lets our BGP Sessions flap. We are currently evaluating the problem together with a Juniper consultant.
Update 15.02.2020 - 12:18: We are currently evaluating to move all traffic from the current QFX5100 to a MX480 in order to exclude any potential issue with a hardware bug, which relates to the Broadcom Chipset which both QFX5100 and QFX3500 uses.