Second thoughts on the R script average tracks:
Although the algorithm partly doesn't give the most satisfying output I used the script for some long ways which I was to lazy to draw all the way by hand.
One use case were long highways without aerial imagery, several uploaded logfiles and a very rough mapping so far.
Another was a river which banks were mapped but not the central waterway. Locally I merged and split the riverbanks to get a way each for the left and the right bank and exported them as GPX file from JOSM to be used for the calculation.
In both cases I checked the result closely and reworked it slightly before uploading.