so the testbed is somewhat increased
my personal testbed consists of notebook and quad-cpu
on both systems the used binaries and data-trees are identical

all time stamps are in CET
cvs-versions and data used:
cvs20101208.1132 + 1.9.0 + EternalLands-update_191_2.zip
cvs20101211.0327 + 1.9.0 + EternalLands-update_191_2.zip
cvs20101215.0242 + 1.9.0 + el_update_191_rc1.zip
cvs20101216.0158 + 1.9.0 + el_update_191_rc1.zip
cvs20101216.1950 + 1.9.0 + el_update_191_rc1.zip
cvs20101216.2243 + 1.9.0 + el_update_191_rc1.zip

for finja i provide the directories
finja-old which is some older system where the bug is reproducable too
finja-new newer and more powerful where the problem only shows up very rarely and almost the same as on my quad-cpu, the horse walks somewhat delayed