Flexure
Flexure at the CCD is measured at about +/-3 pixels over the whole sky
with the locking screws for the CCD stage in place, or about +/-5
pixels (for 15 micron pixels) without the locking screws. This is a
real problem for spectroscopy with a chip which produces pronounced
interference fringes, since the fringes in the flat field will not
match the fringes in the data unless the flat is done at the same
position as each observed object. You may choose to ignore this
source of error, or make a low frequency map of the flats and shift it
to match the flexure, or not use flat fields at all, or take flats to
match every object, or some other method of your own devising, but it
will in any case require some consideration.
Similarly, the flexure will affect your wavelength calibration, but
this may be corrected for by using skylines to re-determine the zero
point for each observation. If you want maximum accuracy you may wish
to do a line lamp calibration before and after each object
observation. It's a nuisance, but they are fairly fast to do.