[jmcomets]

FWIW, most DBMS have built-in index usage stats, and it's not too difficult to query. In my previous job we had a Postgres health dashboard that notably showed `log10(index size) / index hits`, making it pretty clear when some recently added index was useless.

My opinion is that indexes should be either logical (ex: used for exclusion constraints) or purely for performance (tradeoff space for better QPS). Query patterns change, specs change, so monitoring your DB's performance is the way to go.