`init_columns()` still slow for projects with large number of contributions?

[tschaume]

• implement concise correlated subquery with lookup
• use sampling when initializing from database (schema analysis)