Naoyuki reminded us about cosmetics. I hope we have data. The simple way to identify hot pixels is to median together at least 3 bias-sibtracted darks, scaled if necessary by their exposure times if those are different. Find the 3-sigma clipped standard deviation of this frame. If you have at your disposal in the processing libraries make a median-smoother, make a version of this fram with smoothing boxes maybe 65 pixels on a side (?) to get rid of large-scale gradients and subtract this smoohed frame from the original. (This is probably not necessary). Look at the result.
Identify pixels above about 10 sigma and catalog them

The other kind of defect is parallel traps. Without imaging, they are very hard to localize, though one can find the columns in which they occur by
looking for tails from medium-level flats into the vertical overscan.
Charge-transfer inefficiency will typically result in one high pixel
in a `good' column. A trap will result in a several-pixel tail extending
up the column in which the trap occurs. I do not believe these detectors
have significant traps, but we should look.

They can be localized by the 'pixel-pumping' technique, but we have
not implemented it in the CCD code, though we could if they exist and
we need to.