If a reaction to an addressable event points to an e tag and that’s it, then I agree, that doesn’t make sense. If it points to the a tag as is the norm, then we avoid that problem of orphaned reactions. If it points to an a tag *and* an e tag, then we have an option that doesn’t exist if we use the a tag alone: Alice might choose to interpret a reaction to an outdated version as equivalent to a reaction to the current version; Bob may choose to interpret them as not equivalent. The first interpretation errs on the side of being too lax, the second of being too rigorous. Under standard practice, the first interpretation is the only choice.