OSD continues reporting heartbeat-timeouts due to mark_down vs accept race condition

Issue

• OSDs report heartbeat_check timeout messages repeatedly at irregular time interval
• Once the messages start only a restart of the relevant OSD stops them being logged
• OSD logs show that osd.x received no reply from osd.y on either front or back, first ping sent :

7f26973c1700 -1 osd.x 409046 heartbeat_check: no reply from <IP_ADDR:PORT> osd.y ever on either front or back, first ping sent <TIMESTAMP> (cutoff <TIMESTAMP>)
7f26973c1700 -1 osd.x 409046 heartbeat_check: no reply from <IP_ADDR:PORT> osd.y ever on either front or back, first ping sent <TIMESTAMP> (cutoff <TIMESTAMP>)

• OSD logs with debug setting debug_osd=30 shows the following messages :

7f55b5535700 25 osd.x 409978 handle_osd_ping got reply from osd.y first_tx <TIMESTAMP> last_tx <TIMESTAMP> last_rx_back <TIMESTAMP> -> <TIMESTAMP> last_rx_front 0.000000
7f55b1d66700 10 osd.x 409978 tick_without_osd_lock