Instead of creating one aritfact per issue on this matter, create a catch all for the remaining effort:

• [⏳] O(rows): one query per artifact
• [] O(rows): one query per changeset
• [] O(rows): one query per text field
• [] O(rows): one query per list field without pre-compute (eg. MSB).

On the reference tracker (10 columns, 700 rows) that's another 2700 queries saved.