In the fit file, there is a “0”. Interestingly, the first 3 seconds when connection was lost, there are still non-zero cadence values in the file (99 rpm). This seems to be a pattern. I found another 20 s gap in the same file. Speed an HR are actually increasing, while the power and cadence values are 0. Also note the non-integer HR values. 
EDIT:
They should indeed change the standard for recording data. If the connection to a a sensor drops, there should be a -1 or a string instead of a zero. This would simplify data processing/curation.
