T-Mobile Home Internet Frequently Stops Working

Speedtest.net does not run the testing agains the tower equipment. It establishes a connection to a server that can be in different cities. I live in east TN and I run speed tests against the server equipment in Charlotte, NC as that often provides the best results and when testing I select that target even though it is not offered as the "best" source. I can see a Nashville server selected as the best source but get worse results than going to the out of state server. If I go to one in KY it is usually providing poorer results than the server I find useful as a reference server. I use a reference server. You don't have to but the capabilities of random sources for the tests can be quite different. The routing path to the Charlotte server and the capabilities of the Charlotte location is consistent. I am looking to establish a baseline to have a consistent profile. If you jump about from one server to another it is a crap shoot and you get very different results as there are different variables. Sure the amount of traffic across the backbone routers also comes into play but I still find testing agains a reference source is more realistic as I have results over time and more predictable results.