In OSPF broadcast multivendor network with BFD fast failure detection feature enabled, when Juniper router is configured with full-neighbors-only knob BFD session flap may be observed. The syslog message similar to the following will be generated:

Session 192.168.1.1 (IFL 386) state Up -> Down Up time:4d 19:05 Local reason: AdminDown Remote reason: None local discr = 446 remote discr = 33 tx :300000 rx :300000

JunOS allows to configure 'full-neighbors-only' under [edit protocols (ospf | ospf3) area area-id interface interface-name bfd-liveness-detection] hierarchy.
If in OSPF broadcast network an OSPF router goes down, some vendors send OSPF database description packet with Init bit set. When such packet is received by Juniper router, it moves OSPF session to ExStart state. Changing the state from Full to ExStart in such case is expected as per RFC 2328. If this router has full-neighbors-only knob enabled, BFD session will go down with 'AdminDown' reason.

The confirm scenario, on Juniper router ospf traceoptions should be enabled. From the logs:

OSPF rcvd DbD 192.168.1.2 -> 192.168.1.1 (ge-0/0/0.0 IFL 386 area 0.0.0.0)
 Version 2, length 32, ID 192.168.176.182, area 0.0.0.0
 checksum 0x0, authtype 0
 options 0x42, i 1, m 1, ms 1, r 0, seq 0xa9e8b, mtu 1500
OSPF neighbor 192.168.1.2 (realm ospf-v2 ge-0/0/0 area 0.0.0.0) state changed from Full to ExStart due to SeqMismatch (event reason: I bit was incorrect in Loading/Full state) (nbr helped: 0)

When such condition happened, OSPF will instruct BFD to destroy BFD session:

Received Upstream DestroySession (3) len 16:
  Handle (13) len 4: 0x76d90a00
EVENT : Deleting session
Session 192.168.176.1 (IFL 386) state Up -> Down Up time:00:00:01 Local reason: AdminDown Remote reason: None local discr = 446 remote discr = 33 tx :300000 rx :300000

The described Juniper behavior is expected. If it happens in a multivendor scenario, it's recommended to remove the full-neighbors-only knob from Juniper router configuration.