But can the average person tell the difference between a crystal and a microcrystal? I don't think I could. I know I couldn't tell the difference between a precipitate, phase separation or skin effect.

If all the scientists wanted were the obvious candidates, then yes - this would be a good candidate for distributed human checking. But if that was all they wanted, a really simple line detection filter would spit out the candidates much quicker than humans could.

However, the scientists want more data, hence the longer work units. The first load of work units have already been classified by human experts as a control. So, the scientists may be able to drop filters that don't provide useful results, shortening the work unit length. If the filters are letting false negatives slip through, then they will need further refinement.

It's a process. It isn't like a simple production line.