BTA / Delivery desk
Trading Alert Delivery Failures: A Complete Log
Late, missing, duplicate and cancelled instructions belong beside successful receipts, not outside the denominator.
Observe the full chain
Freeze a consecutive observation window before collecting messages. List the source publication point, selected delivery route, destination and any downstream action. A displayed alert proves something different from a transport acknowledgement. Each record should state its exact event type, original identifier and evidence location so another reader can understand the boundary.
When dispatch information is unavailable, screen observations can still describe what the screen displayed. They cannot prove how many messages the source intended to send. Preserve the distinction between a missed source event and an event you cannot establish existed. A delivery success percentage is not interpretable without a defensible expected-message count.
Do not repair stale instructions with a later chart
An invented message might permit an entry only until a stated deadline. If it arrives after that deadline, log the stale receipt even if the price later looks attractive. The later action is not the original decision opportunity. Where a price range or invalidation condition is absent, mark that policy unknown rather than inventing a universal number of usable seconds.
Link retries to their original instruction and keep cancellations visible. A cancellation that arrives after an order creates a reconciliation question, not an opportunity to remove the original alert. Preserve changes to exits separately from changes to the entry window. These states may require different responsible systems and should not be collapsed into one delivered/not-delivered flag.
Distinguish test coverage from reliability
A synthetic route test can exercise receipt, duplication and stale-message handling without touching a live account. It cannot establish performance during every outage, high-volume burst or market event. Keep the test labels in the payload and in exported evidence. Mark any unexercised failure mode as untested rather than assigning it a pass by implication.
Review complete instructions, late receipts, malformed payloads, duplicates, cancellations and unresolved intervals separately. State which device and route were observed. A cryptographic receipt, a message log and a broker response cover different objects. None should be borrowed to certify a missing stage in the chain.
Original Evidence Packet
| Artifact | Record to retain |
|---|---|
| Expected stream | Declared observation window and source count evidence |
| Failures | Late, missing, duplicate, malformed and cancelled objects |
| Bounded conclusion | Tested stage; untested stages and account boundary |
Continue the Investigation
Source Boundaries
Official references describe published mechanisms. BTA's evidence packets and proposed test procedures do not claim a completed paid-feed or account trial.