I guess the solution for this happening in production would be to use the Best Practice exception handling that Blue Prism trains - and that is inforced by the use of Process Templates (which all developers are recommend to use as the basis for their solutions).
If following those lessons you would have a retry loop on sub-pages that would catch an exception thrown because of a disconnection, tidy everything up (close down running instances of app etc), relaunch the app, and navigate back to the required point before trying again.. I would not expect a session to just terminate if there is a rare issue.