As Ryan pointed out, moving the PSF fit down the pipeline would require changing the data structure of TiledBlob which is awkward. Decided to just limit sources to some range at the start of the main script, so that only sources that are going to be processed are initialized. Problem solved! for now
FYI: I won't be able to make the MANTISSA meeting tomorrow at 10. Might be able to make the 11am Celeste/SC meeting, but I'm not sure I'll have a good internet connection. You guys know what the progress on my end is though: I think we can run at scale with #122 as is. I'm looking at scoring next, but that's going to take some work figuring out SQL queries and possibly needing to cache results.