I would be ready to argue that we shouldn't try to process objects larger than some
limit - assuming that we would somehow record this decision and thus enable easy
identification of such cases and specialized post-processing as part of Level 3.
What angular scale do you have in mind?

When I originally wrote this, I thought the most obvious scale would be the size of a patch, which is approximately the same size as a CCD (4k x 4k, or about 13 arcminutes across). Having thought about it a bit more, that would only be the right scale if the object happened to land exactly in the middle of the patch; the patch overlap area is more relevant, and that's only about 1 arcminute, which doesn't seem large enough as an object size upper limit.

So not only are we not terribly wedded to any of the above patch numbers, I think we'll have to have some special-case processing for large objects on patch boundaries anyway, so we have a fair amount of flexibility here. I'd still say we probably don't want to go much beyond the 13 arcminute number (and if we can make it less I think it will make our lives easier in terms of getting processing to fit in memory), but I don't think there are any specific numbers between 1 arcminute and 13 arcminutes that are well-motivated cutoff points.

Zeljko Ivezic has proposed in email that we adopt a 10 arcminute limit. I'm happy to put that in the text of DPDD but do you have a specific paragraph in mind?

I've reassigned this ticket to Mario Juric.

Jim Bosch's reasoning I think is sound - I'm happy to go with the  10 arcminute limit as suggested by Zeljko Ivezic.  For objects larger than this value, we could add a flag to indicate that they were not processed.  The option of developing special code as part DRP processing would need to be studied - I'm more in favour of deferring this to User Generated 

SST consensus on 10 arcminutes