Problem(s) Encountered:

We encountered the same problem reported by Justin Rupert  on 2017 December 8. We saw the icicles show up in the internal flats  taken at the beginning of the night (with ROI 4k x 1 k), even though the count rate was not high enough to saturate the detector. Running snapclean did not seem to fix the problem. In fact, it was not clear if snapclean was running at all since it did not change the ROI to 4x x 4k as expected. We tried changing the  ROI manually to 4k x 4k and then running snapclean, which gave us a good flat in the 4k x 4k configuration.  Switching back to 4x x 1k brought the icicles back. They seem to occupy a smaller area at shorter exposure times but they are still there. Since the weather was bad, we did not really lose time but we were not able to get useable flats.

Solutions:

Snapclean doesn''t change the ROI,  My experience is that snapclear doesn''t completely fix the icicle effect, which appears whenever you use a sub-region.  If you require completely clean flats, it is possible to use 4k x 4k flats and cut out from them the region that matches your ROI.  (Halpern)

Yeah, my experience with the R4K is that the icicles are a real problem with flats.  Remember that the icicles appear to depend on the total counts rather than the peak counts (unlike saturation), which is why they affect flats and not so much the object exposures.  You have to use shorter exposure times to avoid icicles, which means that the peak count level is fairly low.  The good thing is that fringing is negligible, so there is no need to take spectroscopic flats at the position of the object, and I just take a bunch in the afternoon.

Imaging twilight flats have the same issue, but worse, because there is only a narrow window of sky brightness when you won''t have too many icicles. (Chornock)