AGI Bridge Connection behaviour

I have 2 incoming AGI calls LEG1 and LEG2.
I bridge them together by sending a Bridge Command on the LEG1 AGI connection.
The LEG1 AGI TCP connection at this point is blocked waiting for the bridge to clear (as expected).
My LEG2 AGI TCP connection I have in a “holding loop” keeping the TCP connection up - hoping to potentially use this to continue doing call actions if LEG1 hangs up the call.

LEG1 hangs up.
The bridge AGI command returns on the LEG1 TCP AGI connection.
At this point I was expecting to be able to continue to use the LEG2 TCP AGI connection as LEG2 hasn’t hung up.
However from the point of making the bridge connection the LEG2 TCP AGI connection appears to have been lost for any further use.
Any attempt to use the LEG2 AGI connection before or after the bridge results in “511 Command Not Permitted on a dead channel”.

Is there any way to alter this behaviour as it appears that whatever AGI leg is chosen to make the bridge connection this seems to imply it becomes the controlling leg - where the controlling leg hanging up means both call legs are lost.
If in the scenario above if I hangup LEG2, the “non controlling leg”,the LEG1 AGI connection is available to use once the bridge command returns back to the AGI application.

The behaviour is consistent on v1.8.13 and v13.2