I suggest using the 8-figure start of the “no change” snapshot’s SHA to prune it, as reported in the JSON object returned from backup using the --json switch.
I realise that 16 ** 8 is a daunting figure, 4,294,967,296 to be precise. But what would actually happen if the repo found that two SHAs began with the same 8 figures? The obvious thing there would be to try and find the full snapshot SHA based on other known characteristics of the snapshot you wanted to prune, using restic snapshots.
Seeing as this is not an occurrence one can “fake”, to my knowledge, I just want to know what actually happens in a case like this.