How many images should go at a single locus?

By default one — one clear image per station keeps recall clean.

Overloading a locus with several images invites the same interference that crowding loci does: you arrive and can't tell what's yours. Start strict — one image per locus. Later, advanced users deliberately pack a composite image (e.g. a Person-Action-Object triple counts as "one" fused scene per locus), but that's a single bound picture, not a pile of separate items. When in doubt, spread across more loci rather than stacking.