Understand a bad run

vegaload run -target http://127.0.0.1:8080/widgets -protocol http1 \
  -vus 10 -duration 20s -method POST -body '{"name":"x"}' -out bad.json
vegaload diagnose bad.json

diagnose lists findings in plain words: how many requests failed, whether latency has a long tail, and whether the failures were one burst or spread over the whole run. To get ready-made thresholds from a healthy run, use vegaload diagnose -output json baseline.json > gate.json, then -thresholds gate.json on later runs.