Since the topic how how DM handles saturated sources during alert production is a perennial topic, this ticket can serve to collect technical discussions and science use cases in one place.
A user asking about alerts on saturated sources: https://community.lsst.org/t/identifying-suddenly-saturated-objects-in-the-prompt-pipeline/3440. This in the context of a white paper aiming to measure blazar variability, which I will link to once it becomes public.
Some out of context quotes from a Slack discussion:
Eric Bellm [9:50 AM]
I’ve always maintained that any alerts on saturated objects beyond what current processing gets “for free” is a scope increase. And I have not yet been convinced that the science case is compelling when you consider how many false alerts you’d get
Robert Lupton [11:19 AM]
I think all we need know is that a saturated object has appeared in the difference image where the template isn't saturated. But yes, the SDSS code did (and I expect that the LSST code will) include the flux in the bleed trails in the psf and aperture mags. If you assume that the photoelectrons all stay in the buried channel (probably true) then they aren't lost until the bleeds reach the bleed stop/serials. So then you just need to trust the non-linearity corrections to get the total flux
Robert Lupton [11:29 AM]
I think that anything that's now saturated that isn't saturated in the template should be alerted on. In SDSS I set a bit saying "missing flux" along with the bit that said saturated. The flux should still be a lower limit, and caveat emptor. Close pairs of saturated sources, where the bleeds merged, were also labelled bad.
Eric Bellm [11:54 AM]
Do you mean only alerting on saturated things that have counterparts in the template? So saturated things that appear out of nowhere don’t get alerted? And presumably only counterparts that are non-saturated stars, and not galaxies? This decision tree is much more complex than anything else we do with alerts.
Robert Lupton [11:56 AM]
I think all you'd do is to detect on the difference image. If the template is saturated, ignore the alert (or maybe put it out labelled `SATURATED_TEMPLATE`). Otherwise, put it out as a regular alert with (maybe) substandard photometry
Eric Bellm [11:59 AM]
So long as those alerts are not subject to our completeness/purity requirements, and we can actually make some kind of flux measurement, fine. But I still think the scientific value is wildly overestimated. There will be plenty of shallower imaging surveys with better cadence than LSST covering the same sky at the same time.