Thanks for that clarification.

The argument I made in the previous post was more directed to the future scenario when that format change had already happened making such aggregation techniques possible. In that situation, if I understand it correctly, the disproportionate weight of (new) arbitrary data in Taproot envelopes would become evident: We could say that the witness discount then for "normal" signatures/public keys would be justified because they then occupy much less space, but the Taproot envelopes are "cheating" in some way as they can't be aggregated.

Using OP_RETURN would not save additional space in that aggregation scenario, but as the data has no witness discount it would also not be "cheating" in this fashion. (And large Taproot envelopes do need some additional bytes due to the push opcodes necessary to "extend" the data field).