For supernovae, the spectra extraction is more complicated than long slit spectrum extraction because the dispersed spectra often go across the host galaxy and their can be other sources or spectra which overlap.
For most of the mission, the UVOT grisms were only calibrated for a certain part of the detector, so we usually did what was called a "slew in place". This involves targeting the object and then immediately retargeting the object with the grism so that the target was in the right place. This can only be done in the planned timeline so there was usually a couple of days delay to schedule it. Paul Kuin's new grism calibration is valid over a larger area, so grism observations no longer require a slew-in-place in a planned schedule. So they can be done as a rapid ToO upload. Since we did the slew-in-place procedure for so many years (and still do it for grism observations put into the plan) you should probably make it clear when you submit the ToO request that you want the grism observations as soon as possible and don't need a slew-in-place. But for observations scheduled into the plan, a slew in place is probably best.
Here is some of the grism documentation:
There is an IDL program called simgrism by Wayne Landsman which allows you to see how a given roll angle will affect the location of the spectrum and how it might overlap. It shows you what a particular roll angle will look like as far as contamination goes, but you have to know which roll angles are observable anyway. There is a range of about 20 degrees available on any given date.
As far as contamination goes, my grad student Mike Smitka has developed a method to do something similar to galaxy subtraction with the grism images. It isn't a direct subtraction, but uses the template image for the background so it can estimate the non-linear corrections. He describes the method in this paper:
http://adsabs.harvard.edu/abs/2016PASP..128c4501S
Paul Kuin implemented it into the grism reduction software as reading the background from the template image rather than an actual subtraction. But it requires template images taken an integer number of years after the original observation (so that the roll angle is available again). Mike obtained template images for a sample of Ia grism spectra.
Yen-Chen Pan and Ryan Foley have also modified the background subtraction to do better than the default extraction. http://adsabs.harvard.edu/abs/2018MNRAS.479..517P
http://adsabs.harvard.edu/abs/2015MNRAS.449.2514K
http://adsabs.harvard.edu/abs/2014ascl.soft10004K
http://www.mssl.ucl.ac.uk/~npmk/Grism/uvotpy_doc/index.html
http://idlastro.gsfc.nasa.gov/ftp/landsman/simgrism/
There is now an online tool for finding out what roll angles are available for a given ra, dec, and observation date. You can use this with the idl program simgrism.pro to pick roll angles which avoid bright stars in the dispersed spectrum.
https://www.swift.psu.edu/roll.php
Other documentation:
https://swift.gsfc.nasa.gov/analysis/uvot_ugrism.html
https://swift.gsfc.nasa.gov/analysis/uvot_vgrism.html
