does anyone know what is causing media delay after the call is connected in this case?

Usually delays in WebRTC are caused by ICE candidate gathering and I am seeing that on my client (JsSip demo).  However there is another delay in getting media after the call is connected.  

I found an old Jira ticket where Anthony mentioned that on FS demo servers there is a 2.5 sec delay in answering but I could not find info on what is causing this delay.  

It seems to me that FS does not need to gather ICE candidates since it can (and does) take the only candidate from the ext-rtp-ip variable.

So, why the delay?

Thank you

Oleg