I've done various tests using dome flats and worked out a shutter
pattern correction.  Basically, it went like this:

1. I observe both normal (single exposure) flats and multiple
exposure flats of the same requested exposure time.  I was
using mostly a 16-second g-band flat and a flat that was
16 consecutive 1-second exposures before reading out.  This 
gave the exposures comparable flux levels and made some calculations
easy.  The script I used is on flwo48 at:
/home/4obsusers/kepler/scripts/repflat2

2. Suppose an image has the flux F, which is a 2-D function
across the chip.  In a perfect world, F is proportional to the 
intensity I and time t, F=It, where t is what's requested by the 
observer and written in the header.  In reality, t = t0 + delta, 
where t0 is the requested exposure time (a constant across the 
chip) and delta is the residual 2-D function due to the finite 
shutter speed and possible offset from the requested exposure 
time.  If you know delta, you can correct F by the factor
t0/(t0+delta) so that you really have the flux for your
requested exposure time.

3. Using the flats described in point 1, you can form the ratio 
image R by dividing your multi-exposure flat by the single
16-second normal flat.  The flat intensities should be
quite close and drop out:
R = (N(t1+delta))/(t2+delta).  
Here, N=16, t1=1sec, t2=16sec, and the pixel values in R are 
dimensionless.

You can then create an image "delta" from R:

delta = (Rt2 - Nt1)/(N - R)
where delta is in units of seconds.

4. I put a delta image for Keplercam on flwo48 at:
/home/observer/kepler/scratch/delta_kepler_observed.fits
(this is a single-extension fits file assembled from the 4
in the original data files). The image shows a clear 
shutter pattern.

5. In practice, I created a smoothed version of the observed
delta using IRAF's image surface fitting task.  I grab t0 from each
image header and mulitply the images by t0/(t0+delta).

The image shows that the difference in exposure time between
the corners of the field and center is about 0.045 seconds.
That's some combination of the opening and shutting speed.

Mark