This article explains the reason for BGP flapping on backup RE when GRES and NSR are enabled.

If BGP is running with GRES/NSR is enabled, after restarting routing on backup RE or rebooting backup RE or deactivate/activate GRES/NSR, the following “no route table” logs could be seen as routing table hasn’t populated completely and jsr replication has started.

The BGP peer may keep flapping on backup RE with event JsrUnreplicate which means that kernel has closed replication socket for particular BGP session . "show task replication", might show BGP replication complete. "show bgp replication output will give details of BGP replication.  

Master:
Oct  8 02:56:22  marrow_RE0 re1 marrow_RE1 rpd[54320]: RPD_BGP_NEIGHBOR_STATE_CHANGED: BGP peer 10.16.8.2 (Internal AS 65507) changed state from Established to Idle (event JsrUnreplicate) (instance master)
Oct  8 02:56:22  marrow kernel: jsr_psrm_recv_ack_msg(): jsr_psrm_recv_ack_msg(): No route table for our socket w/handle 0x100000100000000 exists on the sec
Oct  8 02:56:22  marrow kernel: jsr_psrm_recv_ack_msg(): jsr_psrm_recv_ack_msg(): No route table for our socket w/handle 0x100000300000002 exists on the sec
Oct  8 02:56:22  marrow kernel: jsr_psrm_recv_ack_msg(): jsr_psrm_recv_ack_msg(): No route table for our socket w/handle 0x100000400000003 exists on the sec
 
Backup:
Oct  8 02:56:22  marrow1 kernel: jsr_psrm_recv_replicate_msg(): no route table for inpcb from replicate message
Oct  8 02:56:22  marrow1 kernel: jsr_psrm_recv_replicate_msg(): no route table for inpcb from replicate message
Oct  8 02:56:22  marrow1 kernel: jsr_psrm_recv_replicate_msg(): no route table for inpcb from replicate message
 

To restore the BGP seesion on backup RE below POA can be executed in a MW.

First perform RE switchover and see if issue resolves if not Disable NSR/GRES and perform RE switchover.