I am glad we are starting to see some movement with the firmware! 
Hopefully the Arduino developers will still provide (shortly) the project files and / or instructions for building the firmware.
Although the issue was written against the server mode, my comments [listed above and in the github issue] reflect that the disconnect problem will happen (in client mode) when tcp_poll_retries is > 8 (so 9) in atcp_poll_conn.
So, I expect the problem with idle disconnect is still present with the client mode of operation.
I will not be able to test this until much later today (at the earliest).
If anyone has a chance to test the client mode for this issue on the new firmware, please do so.
Thanks!