See, this is something that SE should be able to control. Basically from what I've read, if I'm understanding it right, all it would take for the 2002 issue to be minimalized or at least mostly fixed would be a logic cycle that instead of a single reconnect try during that 15 minute disconnect, boost the number upwards. Having it be just one seems to the issue. Upping the code to attempt more reconnect chances would seem to be at least a partial answer, here.

Nice find and very interesting read.