Yes, that is correct behavior in EXOS.
Our OSPFv2 code removes the BFD session when it becomes helpers and re-estabilishes the session after graceful restart. This is done to avoid aborting any graceful restart taking place
My customer reported that the OSPF neighbor is not flap with Juniper or Huawei switches while it is flapped with EXOS or Cisco switches..
As hitless failover flag was set during client registration with BFD, BFD session were retained after failover.
After that, OSPF helper router (EXOS x460) deletes BFD session as intiated by OSPF and this makes BFD session at the middle switch which underwent MSM failover ends to go down with admin state DOWN flag.
However forwarding table and LSDB is remained during restart period so there is no traffic loss
during MSM failover.
RFC 5882 section 4.3.2 also seems to recommend that it is best to avoid aborting the graceful restart taking place.