My guess about the problem in 15.30 was that the "auto" barometer watch mode was sometimes incorrectly interpreting altitude changes as air pressure changes. In my case, driving a car seemed to often trigger the problem in 15.30. They definitely tweaked the algorithms in 15.40, since I personally haven't seen it since. But considering how difficult it is to write an algorithm which correctly guesses whether it is the altitude or the air pressure that changes when the watch only has a barometer and an accelerometer, I wouldn't be surprised if the current version has its limitations also.
