madoverlordOne possible thing we could do -- with a lot of help from supporters -- is go over the films frame by frame and do dust removal.
A Galaxy Zoo/Planethunters style crowdsourced artefact-identification seems like an excellent idea in theory (though almost a kickstarter project in itself to set up). Probably not the best idea to crowdsource the actual image editing, but cutting down on the tedious identification process would speed the process up a lot and keep the costs of the editing itself down.
A bit of back-of-the-napkin maths: 335min runtime * 60s * 24fps = 482,400 frames. To get the best results with minimal false positives, every frame would need to be looked at by multiple people. At the current nearly-1,200 backer level, and with say 3 viewers per frame, each backer would need to look at ~1200 frames for full coverage. With more backers, and the general public involved, checking the entire running is definitely feasible, especially with a few dedicated fans going above the average.
It'd be a little hard on bandwidth though: 482,400 * 3 views * 2.2mb per image (assuming 24bpp and PNG comrpession ratio of 2.7:1) = a bit over 3tb of transfers (and 1tb of storage). That would probably need to be tripled to allow for a 3-frame flipbook to aid in identification ("is that a smudge that appears in one frame, or a bit of background art that persists over multuple frames?"). 9TB is still completely doable.
Again though, it'd mean dumping the existing JP BD encode and re-doing a whole lot of work.
Its definitely something to pursue for cases where a HD master needs to be created (or a scan exists but existing encodes are poor) and could drop the cost of doing basic cleanup of older footage. Especially if as well as just identifying and locating damage, a second round was conducted for triage in order to identify easily fixed 'low hanging fruit' damage (e.g. in dark areas, blank frames, frames with no background motion) to allow best-bang-for-your-buck fixing.