> Due to the hash functions being one-way, we cannot derive your data back from the hash
This isn't true in practice. It's not hard to build a big list of ~every email address (give or take), and have a GPU churn through them all until you get a match.
If you've ever received a spam email, your email address is on such a list.
argon2id with dynamic salt should effectively prevent this, but it will also not allow to tell if two users have the same e-mail or not -- which I suspect is the main reason for hashing in the first place.
If equality-check is required to prevent e-mail reuse by spammers then argon2id with static salt rotated every few months will be reasonably strong too.
Of course I have no idea if any of this is implemented or it's just sha256(email). Just replying to the question of general feasibility.
