The suggestion was made that the Read1 sources detected by FREXAS be passed on by POSFRM for use in MAPCOR even when they have fewer than 6 apparitions. That way MAPCOR could use them to search for persistence in the scan, although such sources at the end of a scan would have to be carried forward for use in the next scan. Unfortunately, this looks impossible with the current asynchronous scan processing design (i.e., a scan needing such Read1 sources from its predecessor could easily be processed before it).
It would be possible to handle Read1 sources at the beginning of a scan this way; the effects due to Read1 sources at the end could be left for final catalog preparation time, since some cleanup will be required then anyway. The incompletely covered Read1 sources would be appropriately flagged so that BANDMERGE and other subsystems could take appropriate action (TBD).
It was pointed out that another approach to persistence could also be implemented that uses the identification of pixels on which the bright source was observed. Currently the search for persistence objects is done in space coordinates rather than pixel coordinates. Since the pixel hit by a bright source would be expected to decay exponentially, a search using a predictable pattern of negative Read2-Read1 sources involving that pixel should be efficient. It is not clear why the Read2-Read1 subtraction does not always produce negative sources, which would eliminate persistence objects, but apparently the persistence behavior is erratic enough to create some positive artifacts. It is also not clear how the poor fit involving a single frame persistence object stacked up with several apparitions of noise still emerges from KAMPHOT as a point source; some chi-square tuning may be needed, and perhaps these objects are systematically biased toward having some empty apparitions rejected because of dead pixels.
The consensus was that the passing on of incompletely covered Read1 sources at the beginning of scans should be done, with changes to MAPCOR (and maybe PROPHOT) to use (or avoid) the information, and that BANDMERGE must have general purpose flag-dependent processing control. Furthermore, the pixel space search method should be implemented (presumably in MAPCOR), and global (i.e., not scan limited) persistence searches must be planned for final catalog preparation.
It was pointed out that persistence could be caused by bright objects observed during slews, since there is no shutter. The cost of waiting for persistence to fade before beginning a scan has been investigated already, and it has a surprisingly large cost in survey time (reasonable delay times lead to survey extensions of about four months). Besides hoping that the slew rate will be sufficient to prevent serious persistence problems from slewing over bright objects, no plan for cleaning up such problems came forth.
The team preference seemed to be to provide the images with warning symbols placed at artifact locations, and to provide information in the FITS headers that could be used by IRAF and Skyview to replace these symbols with the original pixels. Users would thus be reliably made aware of the artifacts without requiring great effort to get at the original pixels. Where the symbol insertion and FITS header expansion would be done was left open.