What share of a warehouse's tables are never read after being written, and how did teams find out?

Open question

Every pipeline that is built adds tables; few pipelines are deleted when the report they fed is retired. The data to answer "is anyone reading this?" exists in most engines: the PostgreSQL documentation describes pg_stat_all_tables with seq_scan and idx_scan counters and, in current versions, last_seq_scan and last_idx_scan timestamps per table, and warehouse products keep query histories that can be joined to table names. Yet the wiki has no record of what teams found when they actually ran that analysis: what share of tables had not been read in a month, a quarter or a year; whether unread tables clustered by age, by owner, by pipeline layer (raw, staging, marts) or by the person who left; and how much storage and pipeline runtime they consumed. There is also a measurement problem: a table read once a year by an audit or a year-end report looks unread for eleven months, and reads by the pipeline itself (staging to mart) must be separated from reads by people and applications. Which analyses handled this, and how? Finally, what did teams do with the answer: delete, archive, stop the pipeline but keep the data, or mark the table deprecated and wait for complaints, and how many complaints came?

What a useful answer contains

The engine and how reads were attributed (statistics views, query log parsing, lineage tooling), and the observation window. The number of tables, the share unread in the window, and the breakdown by layer, age and owner where available. How pipeline-internal reads were excluded and how rare but legitimate readers were identified. The storage and compute attributed to unread tables, if measured, with the method. The action taken, the notice period given, the number of objections received and how many tables were restored. Whether the analysis was repeated later and the share changed, and whether a rule (for example, a table with no external reads for a defined period is archived) was adopted as a result. One-off findings are welcome if the window and the exclusion rules are stated.