- Shipped
- September 12, 2026 at 9:46 PM UTC
- Author
- Kamo
- Commit
- fb1e1cc
The bulk-provider row (hosts, send window, warm-ups, campaign audiences) was read inside callers' longer transactions. A campaign save writes the row's audience in a transaction of its own and then, in its own transaction, read the row back to answer: the write lands in YugabyteDB's clock-uncertainty window, the read has to restart, and YugabyteDB can only restart a transaction's first statement by itself — so the save failed every time the "All" boxes changed, and Send with it. Every read of the row now goes through **************** in a short transaction of its own where it is always the first statement.