Live 2026-08-23 CURSOR review
GET https://congress.trade/api/transactions?order=desc&limit=8 returns Ro Khanna 2024-03/04 PTR rows (H-2024-8220192, source=local_mac, firstSeenAt=2026-08-10) as if they were the newest trades.
Tests in app/src/delivery/__tests__/buildTransactionsQuery.test.ts document that omitted/desc order is ORDER BY t.cursor_seq, not tx_date / first_seen_at. A large historical backfill with high cursor ids wins the public "latest" snapshot.
The web Trades tab with the default 3-month window is fine (Ed Case AAPL 2026-08-13 on page 1). API clients, Delivery consumers, and any iOS path that uses unwindowed order=desc are not.
Fix
Default public newest-first to COALESCE(first_seen_at, filed_date, tx_date) (or document sort=tx_date and make the web/iOS/docs all use it). Add a contract test that a 2024 backfill row cannot outrank a 2026 primary row on order=desc.
Keepouts: extract/halt, Infisical.
Live 2026-08-23 CURSOR review
GET https://congress.trade/api/transactions?order=desc&limit=8returns Ro Khanna 2024-03/04 PTR rows (H-2024-8220192,source=local_mac,firstSeenAt=2026-08-10) as if they were the newest trades.Tests in
app/src/delivery/__tests__/buildTransactionsQuery.test.tsdocument that omitted/descorder isORDER BY t.cursor_seq, nottx_date/first_seen_at. A large historical backfill with high cursor ids wins the public "latest" snapshot.The web Trades tab with the default 3-month window is fine (Ed Case AAPL 2026-08-13 on page 1). API clients, Delivery consumers, and any iOS path that uses unwindowed
order=descare not.Fix
Default public newest-first to
COALESCE(first_seen_at, filed_date, tx_date)(or documentsort=tx_dateand make the web/iOS/docs all use it). Add a contract test that a 2024 backfill row cannot outrank a 2026 primary row onorder=desc.Keepouts: extract/halt, Infisical.