Why not use a bitemporal data model instead, it's cleaner and easier to reason about.

https://en.wikipedia.org/wiki/Temporal_database

I think it would be the "naive" implementation. For specific data structures more efficient specific algorithms can be written.

You can also do time-travelling control structures: https://hackage.haskell.org/package/tardis-0.3.0.0/docs/Cont...

Perl did time-travelling variables first, namely in 2007: Positronic::Variables

https://www.flickr.com/photos/gunnarwolf/1343377956/

This in the picture is Conway giving the talk.

Can anyone give an example of a use case for this?

I'd be rather curious; this is an interesting concept.

Not sure I understand the problem. If each submission is interesting in and of themselves, does it matter if the source is the same?