Here's how often we fail.
Real processing outcomes from the last 30 days, including our own test runs. Completion is not a guarantee of accuracy.
Live reliability readings load when you open this page.
Success rate = completed ÷ (completed + failed). Only retained tasks submitted in the last 30 days are counted; queued and running tasks are excluded. Restarts replace earlier attempts, and deleted tasks are absent. Median time covers completed tasks with valid start/end times, excluding queue time. Readings are cached for five minutes; reload for a newer snapshot. Flux counts conversion jobs; a batch is completed when at least one file succeeds. Flux failures are unclassified. Axis counts successful and failed coordinate-transform requests started in the last 30 days, once per request including batches, from the tracking start shown below. Earlier success-only history is excluded. Axis HTTP 4xx responses are input rejections; HTTP 5xx and unreachable services are system failures. Flux runtime is not recorded.
Worker availability is a separate recent worker observation, not an end-to-end test or an account entitlement. Quality-check probes are not available. Partial means only one of the two positioning workers is ready; Apex Duo requires both. Unknown means no supported, recent observation. Worker observation: