Interesting! I've seen your posts on this subject, gerhard, and you obviously really know your stuff.
I do hope that GlobalSat fine-tune the distance measurement. You could be right that the difference is that they now have greater confidence that the trackpoints recorded are accurate (because of the superiority of the SirfStarIV chipset), and that they can thus 'join the dots' without applying aggressive error correction. Is that what you think is now happening? I'll have to try the watch out on one of my more winding trails, where I could see the average pace dropping with the 615B, and see if the same thing happens with the 625-XT.
I don't know whether this is already possible, as I haven't explored some of the new features of my watch, but it strikes me that it would make very good sense to have activity-dependent error correction. So, if you set the watch to 'Trail Running' or 'Walking', it allows for the fact that the trackpoints may legitimately zigzag, as by the nature of the activity you do often have rapid changes of direction, whereas for a 'Cycling' setting, it would know that as you were going faster and in a straighter line, points recorded far out from that line are probably errors.