Just to add one tiny thing to this excellent post! IIRC, you can, in fact, determine the PSF quite well ... on an individual plate (or region of a plate), but there is considerable variation between plates (certainly epochs!), and (possibly, I just don't remember) poorly constrained variation between different parts of the same plate.

The digitisation of the original plates was done quite well (the plate machines are - were? I don't know if they're still in use - good examples of precision engineering), but the project's objectives did not include producing a product from which consistent, accurate, well-defined PSFs could be extracted*.

If anyone's interested, since the DSS data is in the public domain, you could have a go at determining the PSFs for yourself ... there are certainly plenty of 'unresolved, point sources' (i.e. stars) all over most plates ... which you can find from one or more of the many online databases.

*At least I don't think the objectives included this; I could well be wrong though.