When Event-Driven Systems Become the Source of Truth: Lessons from CDC, Kafka, and Replay-Safe Inventory Platforms
Abstract
This paper examines what changes when CDC and Kafka streams become the operational truth for inventory state. It draws on a production inventory platform and an anonymized recovery of approximately one million events. Redis provided short-window duplicate suppression, while DynamoDB retained longer-lived processing evidence. The recovery pipeline and runbook described here were implemented and used in production. Their identifiers and labels are illustrative. The paper traces how ambiguous outcomes, offset resets, ordering mistakes, and stale projections can turn replay into another source of error. It separates message delivery from inventory mutation, then explains the controls needed to repair state: stable business-effect identity, bounded replay, audit records, and reconciliation against authoritative history. The practical test is whether a team can establish what happened, control the scope of replay, and prove the resulting business state is correct.
- Published in
- Proceedings of the Raptors Conference, Vol. 1 (2026)
- Conference
- Raptors Conference 2026, 21 October 2026, online
- Accepted
- 2026-09-19
- DOI
- https://doi.org/10.68495/rc.2026.C028
- Presentation
- Slides (PDF)
- Licence
- © 2026Ishan Deepak Shah. Published under Creative Commons Attribution 4.0 International (CC BY 4.0).
- Publisher
- Hackathon Raptors, Community Interest Company 15557917, London, United Kingdom
- Permanent address
- https://proceedings.raptors.dev/shah-when-event-driven-systems
Open access. Every entry in the Proceedings is free to read, with no charge to authors and no charge to readers. Slides are served from this site, so they stay available at the paper's permanent address.
Cite this paper
Ishan Deepak Shah (2026). When Event-Driven Systems Become the Source of Truth: Lessons from CDC, Kafka, and Replay-Safe Inventory Platforms. Proceedings of the Raptors Conference, Vol. 1 (2026). Hackathon Raptors, London. https://doi.org/10.68495/rc.2026.C028
@inproceedings{raptors2026-c028,
title = {{When Event-Driven Systems Become the Source of Truth: Lessons from CDC, Kafka, and Replay-Safe Inventory Platforms}},
author = {Shah, Ishan Deepak},
booktitle = {Proceedings of the Raptors Conference},
volume = {1},
year = {2026},
publisher = {Hackathon Raptors},
address = {London, United Kingdom},
doi = {10.68495/rc.2026.C028},
url = {https://doi.org/10.68495/rc.2026.C028},
abstract = {This paper examines what changes when CDC and Kafka streams become the operational truth for inventory state. It draws on a production inventory platform and an anonymized recovery of approximately one million events. Redis provided short-window duplicate suppression, while DynamoDB retained longer-lived processing evidence. The recovery pipeline and runbook described here were implemented and used in production. Their identifiers and labels are illustrative. The paper traces how ambiguous outcomes, offset resets, ordering mistakes, and stale projections can turn replay into another source of error. It separates message delivery from inventory mutation, then explains the controls needed to repair state: stable business-effect identity, bounded replay, audit records, and reconciliation against authoritative history. The practical test is whether a team can establish what happened, control the scope of replay, and prove the resulting business state is correct.}
}
TY - CONF TI - When Event-Driven Systems Become the Source of Truth: Lessons from CDC, Kafka, and Replay-Safe Inventory Platforms AU - Shah, Ishan Deepak T2 - Proceedings of the Raptors Conference VL - 1 PY - 2026 DA - 2026/09/19 PB - Hackathon Raptors CY - London, United Kingdom DO - 10.68495/rc.2026.C028 UR - https://doi.org/10.68495/rc.2026.C028 AB - This paper examines what changes when CDC and Kafka streams become the operational truth for inventory state. It draws on a production inventory platform and an anonymized recovery of approximately one million events. Redis provided short-window duplicate suppression, while DynamoDB retained longer-lived processing evidence. The recovery pipeline and runbook described here were implemented and used in production. Their identifiers and labels are illustrative. The paper traces how ambiguous outcomes, offset resets, ordering mistakes, and stale projections can turn replay into another source of error. It separates message delivery from inventory mutation, then explains the controls needed to repair state: stable business-effect identity, bounded replay, audit records, and reconciliation against authoritative history. The practical test is whether a team can establish what happened, control the scope of replay, and prove the resulting business state is correct. LA - en ER -