Description
Field testing with buggy gateways and intermittent networks indicates that link-tracking is fragile. If it gets to the point of wanting to rejoin, various things can happen, but generally things lock up.
The change that used OP_REJOIN to send an occasional join request simply doesn't work. In addition, although forcing a join at when we get to the end of our rope is not bad, because of issue #2, and because of the "dead gateway downlink" problem in TTN US, we really want to be sparing with triggering joins. Link tracking should not trigger a join until a much bigger count (unless, of course, the application decides it's a good idea). By much bigger, I mean thousands of uplinks without downlinks.
Perhaps once I replace engineUpdate() with a proper FSM, many of these things will be easier...