Bug: when filtering Workflow Runs by status "success", some runs are omitted

This appears to happen both through the Github API using the “List Workflow Runs” call, as well as in the Github UI.

The list of all runs for this workflow shows that there are a total of 11 runs of the “Workflow Runs Create Test Data” workflow.

Counting by hand shows that there are 8 runs that completed with a status of “success”: Run Numbers 1, 3, 5, 6, 8, 9, 10, 11.

Likewise, using the github api shows the same, when all runs are requested and then post-filtered using jq:

gh api /repos/bantic/github-metrics-tests/actions/workflows/source-of-run-data.yml/runs \
  | jq '.workflow_runs[] | select(.conclusion == "success") | .run_number'
11
10
9
8
6
5
3
1

However, in the Github UI when selecting “success” from the “Status” dropdown, only 6 (not 8) results are shown. The successful run numbers 5 and 6 are omitted.

Likewise, using the API results in the same:

gh api /repos/bantic/github-metrics-tests/actions/workflows/source-of-run-data.yml/runs \
  -F status=success --method=GET \
  | jq '.workflow_runs[] | .run_number'
11
10
9
8
3
1

I had written some integration tests using this API call several months ago and recorded Github’s API responses as fixtures in the test suite, so I can confirm that in approximately May 2021 the API call with the status=success filter did return all 8 successful workflow runs.

There doesn’t seem to be anything different about Run 5 and Run 6 than the other successful runs. Why are they not included in the API call?