From Fast Runs to Inspectable Evidence
Cloud Health Office moved from repeatable throughput runs into broader edge-case scoring and operator-visible evidence: expected pends are observed from persisted claim status, unsupported scenarios are separated from mismatches, and the Mass Adjudication console can inspect run summaries and claim-level evidence. The latest published scale result is one million mixed claims through Azure Service Bus at 155.89 claims per second, plus a separate 100,000-claim raw X12 837P run at 199.42 end-to-end claims per second. Both are local Kubernetes validations on Docker Desktop, not production cloud benchmarks.
The asynchronous validator observed 999,878 claims within its 180-second windows and timed out on 122 observations clustered in two brief episodes. Post-run reconciliation found exactly one million terminal claims in MongoDB, with no dead letters, pod restarts, or error logs. The separate raw 837P run accepted and adjudicated all 100,000 transactions.
Local Docker Desktop Kubernetes benchmark, not a production cloud benchmark. Read the methodology · Read Part 16 · Engineering series · Evidence archive →