I was actually thinking about dynamic forces with respect to Matt's oil migration theory the other day, but after a very rough back-of-an-envelope calculation realised that dynamic effects caused by the movement of the mechanism itself should have little influence on lubricant distribution or migration.
If I estimate the escape wheel to be 5.0mm in diameter and rotating at the rate equivalent to 3.5 teeth per second, this equates to an average rotational speed of 2.75 rad/s (it has 8 teeth). Now this is of course a drastic over simplication, as the escape wheel is repeatedly starting and stopping, and not rotating at a constant rate. The molecule of Moebius therefore experiences a jerk and much higher acceleration. However considering that 0.02m/s^2 is several orders of magnitude smaller than acceleration due to gravity, this leads me to conclude that the effect of the rotation of the escape wheel is not such that the lubricant will be redistributed. After all, the viscous shear forces within the oil also need to be overcome, and as the escape wheel rotates gravity alone would have a higher tendencey to re-distribute oil than the dynamic motion alone. In short, the escape wheel is too small and rotating too slowly to cause the lubricant to move around.
Anything more than this rough calculation would require me to reach for textbooks I haven't touched for many years and it's also a while since I did anything in ANSYS...
The idea proposed by Matt of small channels on the surface of components (as a result of by imperfections, contaminants, burr) causing lubricant to migrate through capillary action (think about WD40 creeping into grooves and holes when it is applied to a machined component) is rather plausible.
So, let's see if I can't turn this into the most boring thread on the forum