Fix how my app handles large data exports (CSV downloads, full-account data exports, admin reports). Find every export path and check what happens when it covers 100,000 rows instead of 100. Apply the right pattern per size: small exports (bounded, a few thousand rows) can stay synchronous with a query timeout; anything larger moves to a background job that generates the file in chunks (cursor or keyset iteration, never loading the whole result into memory), writes it to storage, and gives the user a download link or an email when it's ready. Stream responses where synchronous generation stays: write rows as they're read instead of building one giant string. Cap export size and rate limit export requests per user, since exports are expensive and abusable. Don't leave the current 'load everything and stringify it' approach in place behind a loading spinner; that's the thing being fixed. Don't break the file format users already get; same columns, same encoding, and make sure values with commas and quotes escape correctly. Deliver: the per-export pattern chosen, the streaming or background implementation, the size caps, and verification: run an export over a seeded 100,000-row dataset and confirm memory stays flat and the file is complete and well-formed.