If they want to track runserver calls, they presumably realize that there will be frequent runserver calls in the context of developing/testing an application. I would think they would care more about the number of unique IP addresses those runserver 'runs' are coming from.
The exact scale would be incorrect, but since counting errors can be expected to happen the same way for other packages in the same space, it's still enough to see relative popularity, growth in usage, etc.