If I had to place a bet on the availability 100 or 300 years from now of a digital photo on the Internet, versus a preserved specimen in a natural history museum, I know where my money would be going (having worked extensively with both).
And if we are talking physical photos, there’s really not much practical difference between conserving those and conserving the specimen from which they came, so why not retain the primary evidence instead of (or along with) secondary derivations?
I do agree that vouchering can be overdone, and should not be. (After all, I’m here on iNaturalist, and believing wholeheartedly in the value it is adding to human knowledge and interaction.) But the more frequent cases I have noted are when vouchering is underdone, rendering impossible the repeatability of a scientific study.