Falko Menge

Could it be that you had unfinished process instances from an earlier run inside the engine that now got finished as well?

still ;-)

+1 In fact, the transaction boundaries configurable with `asyncBefore` and `asyncAfter` are the only ones that are really before or after the flow node. The implicit TX boundaries of wait...

The original code is available here: https://github.com/camunda/camunda-bpm-process-test-coverage/blob/master/core/src/main/resources/bpmn.js-report-template.html

@xomiamoore FYI

FYI @jothikiruthika

The Opensearch Exporter is likely affected by the same issue as its [index templates](https://github.com/camunda/zeebe/tree/main/zeebe/exporters/opensearch-exporter/src/main/resources) also have inconsistent numbers of shards.

@remcowesterhoud cross-posting my comment from the GDoc for visibility: > The feature has been requested specifically for a high-performance use case. Which impact will the use of the REST API...

Could we emit a warning if the values don't match?

@chDame FYI