Hairpins go on a hike

This is an issue that keeps occurring to me, but which I can't describe with adequate precision to file a bug report or feature request. I am working with 2.0.3. and MacOS Yosemite.

This usually only happens in large scores (say a 10 minute string quartet movement or an orchestral score). After repeated work on a score (1. typeset, 2. proofread and correct, 3. format, 4. generate parts, the format parts for page turns etc., all the while correcting errors one missed on the first proofread). I find hairpins that have moved horizontally away from their anchors, mostly but not always forward. Not all hairpins in a score move and not all of those that do move move the same distance. Sometimes it is obvious that a hairpin is in a nonsensical location, but often it isn't. The only way to be sure to catch all hairpins that have moved is to double click each of them and have a look at the anchor lines (correcting this is obviously easy). I remember instances with other line elements like Pedal (piano) but I am not sure if these occurred with the present version of MS.

Here are two hairpins at the same place, moved one forward the other backward (BTW: MS's built in screen shot facility does not allow to take this picture, you'd have to unselect the hairpins to take a--useless--picture):

As I said I know this occurs in large scores that have been worked on for a lot of time (which includes numerous save/load cycles), but I have no observations of details that would be connected to this behavior. So I am just asking some questions to the community:
- Has anybody encountered this behavior?
- Any observations about when this occurs, what triggers it?
- Has anybody got far enough to find a way to prevent it? Or can tell me where I am going wrong and cause it myself?

This is obviously a labor intensive problem as one needs to check on and correct every hairpin before printing.