Every place the order id is held as 32 bits, across a four-million-line monorepo
31 found across 9 services, 7 unconfirmed, 214 not read. 46 grep hits ruled out, among them purchase-order ids and a sort_order_id.
A scenario, not a customer: commit a41c9e0, 63,870 files in 14 directories of Java, Kotlin, Python, Go, TypeScript and SQL. Sent at 10:02:05, answered at 10:26:31.
Java services - 11: six DTOs, two parseInt calls, three retry-queue casts
Reporting job - 6, four of them CAST(order_id AS INT)
Protobuf - 4, ShipmentCreated and InvoiceIssued among them
Unconfirmed - reflection twice, the mobile app's repository, two vendors, a deploy-time mapping, one templated SQL file

Three more questions of the same shape as the order id
Ask one of your own


Every-place means one value that starts somewhere and travels through services, formats and languages, where the answer needed is the whole list, not the best match.
Ask which place is a bug and it declines, without billing you
A guess would read exactly like an answer. A declined question costs nothing and comes back with an every-place question that covers part of it.
Which of these 31 is a bug: declined, as a judgment, not a place. It offers every place the id is compared or sorted instead.
Is it safe to widen ShipmentCreated.order_id: declined. Safety depends on readers outside the snapshot and on deploy order.
What the payment provider does with the order reference happens at runtime, in another company's system. The most it gives is where the value leaves the snapshot.