There are two components that people refer to as "vignetting" in the NICMOS cameras:
The dewar anomaly has pushed the cameras forward and for some reason (probably a tilt between the foreoptics and the dewar, but no one is really sure) we are now imaging the edge of the Field Divider Assembly (FDA) at the bottom of all 3 cameras. The FDA is a fold mirror that sends the light from the foreoptics down each camera channel and through the filters and dewar window to the cameras. The bottom edge of all 3 cameras now looks past the edge of the FDA into the cold stop inside of the NICMOS. This means that there is less signal from the telescope along that edge, and thus the throughput is decreased. This is similar to the effect at the bottom of the WFPC2 cameras which look past the pyramid on those edges.
In NIC3 ~15 rows along the bottom edge are affected, and the throughput there is about 30% of the mean of the rest of the array. For NIC2 it has about the same extent, but is only reduced to 95% of the mean of the rest of the array. NIC1 is even less affected. Note that this hard edge has been moving around with time, and so its extent changes. Motions of the coronographic hole, which is also on the FDA, track the motion of this edge very well.
The entire NICMOS now sees a part of the forward bulkhead of the NICMOS at the entrance aperture of the camera. This warm metal surface emits a large amount of radiation in the longer wavelength bands. This emission extends over about 50 rows in NIC3 and something like 20 rows in NIC2. By moving the Field Offset Mechanism (FOM) we can see a different part of the sky in the HST focal plane, and essentially look farther away from the bulkhead that is causing the emission. This motion is automatic for camera 3 observations after January of 1998. The FOM however is forward of the FDA in the optical path, so it has absolutely no effect on the FDA vignetting (explained in part 1 above).
The effect of setting the nominal NIC3 FOM to Y+16 has completely removed the warm emission (part 2 above), but has not changed the hard vignetting in the bottom ~15 rows (part 1 above) at all. Furthermore, the motion of the FDA relative to the detectors means that the vignetted edge at the bottom is not always in the same place, so the flatfield reference file will not always "correct" for the vignetting by the proper amount. This shows up as a dark or bright edge along the bottom of the calibrated image. It is *NOT* the thermal emission of part 2 above, it is simply an over (or under) correction of the throughput by a flat reference file made when the FDA was at a different location relative to when your data were taken. Typical shifts in the FDA result in throughput changes of order +/- 5% along that bottom edge.