When Queues Undermine Reliability: per-message records

Per-message records and per-second timelines for 99 experimental runs (48 on an EC2 m6i.4xlarge, 51 on a laptop) measuring how RabbitMQ and Apache Kafka behave when a consumer slows down for minutes rather than failing. Each run is two files: a timeline with a provenance header and one row per second, and one row per publish attempt. Also included are the broker logs, metric samples and orchestrator events from which the alarm counts and purge depths in the paper are derived. SHA256SUMS and MANIFEST.csv accompany the data. Code and analysis scripts: https://github.com/violetta98/when-queues-undermine-reliability