The manual of autossh clearly states that if you dont need it for purposes other than checking of the connection broke, you shouldnt be using monitoring.
For example, if you are using a recent version of OpenSSH, you
may wish to explore using the ServerAliveInterval and
ServerAliveCountMax options to have the SSH client exit if it
finds itself no longer connected to the server. In many ways
this may be a better solution than the monitoring port.
So just disable autossh monitoring with
-M 0 and configure the
ServerAlive options (interval 30, max 3 is a reasonable default) instead.
You next debugging step (after ensuring the new sshd config is loaded) is checking the server log (sshd will likely log to
/var/log/auth.log) and a verbose client log (try something like
ssh -vvv -R port:ip:port master on one of your servers and see if the connection loss is indeed not network-related).