This thread was started specifically regarding the question of how to deal with situations where a user edits the observation media in such a way that previous IDs no longer apply to the current media. (I will note here that changes to observation media may affect more than just the IDs, but any of a number of additional elements that might be added by users, such as annotations, observation fields, projects, DQA votes). A suggestion that ensures formal continuity but does not address this core problem thus seems to me to be rather beside the point.
In fact, it seems to me that your proposal might in fact create more such situations rather than reducing their occurrence. Users often struggle with editing photos (deleting different ones than they intended etc.). If they accidentally delete the wrong photos they will then be stuck with a photo that they do not want to be part of the observation, even if they subsequently re-add the correct ones, which is likely to produce frustration (why can I edit all the photos except this one?) and users either giving up and leaving the observation with a non-applicable photo, or deleting the observation and uploading it again, and thereby destroying previous IDer effort and energy. Neither of these outcomes seem desirable.
While it may happen occasionally that users replace their initial photos with photos of something completely different with no apparent motivation for doing so, I imagine this is fairly rare, and generally the result of users not understanding or struggling to use iNaturalist. Why? Because if one wants to completely change the observation, it is typically easier to create a new observation than to edit an existing one. Most of the time when people are editing media, it is to correct a mistake; in many cases this is a mistake that they only notice after it is pointed out to them by another user, who may also have interacted with the observation in the form of adding an ID. In other words, most edits – including situations where people have replaced all media items with unrelated ones – are made in good faith, not with an intention of falsifying previous IDs. Locking elements of an observation feels punitive, would send the message that observers are considered basically not trustworthy, and it would take away some of their agency over their observation, including their ability to correct mistakes.
It would not.
How are you defining continuity and why is it important?
How does preventing deletion of one arbitrarily designated photo (of potentially multiple media items) ensure continuity in a meaningful way?
How do you balance this with the need for observers to be able to edit observations in order to correct mistakes or other issues?
I’m not asking this merely for the sake of argument. I think you are basing your idea on assumptions about identity and continuity that are not helpful for iNaturalist observations, nor do they hold up to logical scrutiny.
There is a philosophical concept known as the “Ship of Theseus Paradox” that is relevant here. Basically this is the question of what conditions are necessary for a thing (whether a ship or an iNaturalist observation) to be considered the “same” thing over time even if various elements are changed or replaced. It is not as straightforward as you seem to believe. What makes one particular photo – which you are defining merely as the photo that remains after any others have been deleted – constitutive of that observation as a particular unique entity? Why that photo and not some other one? Why does replacing a media item with a different version of the same file not represent continuity? Why are location and date not part of the core identity of an observation (given that iNat defines an observation as an interaction with a particular individual at a particular place and time)? If the identity of an observation is defined by the presence of a particular photo, what about observations that have identical media?