If you add a private salt, you won't be able to do "grouping" or identifying duplicates, which is what this thread was discussing.
If it was me, and I wanted "independent" researchers to highlight clusters or duplicates I would do the following as a first-pass solution:
Store an internal mapping of IP->unique sequential number, likewise do the same for usernames. The goal is that it's random and not based on any hash or ordering. So people with either the IP, username or username + IP, can't identify the unique internal numbers.
Then release those. Though tbf, if I was part of any sort of "bot prevention" or "sock puppet identification" team at Reddit, I'd be doing this already. But we all know the dirty secret is to not actually track down such abuse, but to appear like you are doing so, so that you can inflate your user count with plausible deniability.
