I'm running tests on Git for Windows that I compile myself. To do so I'm using the following command line insode the git/t directory:

$ prove -j 5 -a test-results.tar.gz ./t[0-9]*.sh

Please find the generated archive attached.

I'm later unpacking the archive and using this iin the TAP plugin's "Test results" field:

test-results/*.sh

This results in the following page:

https://qa.nest-initiative.org/view/msysGit/job/msysgit-mingwGitDevEnv-test/lastCompletedBuild/tapResults/

As you can see, for all failed (sub-)tests there are lines starting with "#" in the table instead of the number of the failing test and the other columns. As a result, the summary "8641 tests, 8204 ok, 1 not ok, 1 skipped, 0 Bail Out!." at the top of the page is wrong. I've manually extracted the correct summary to this page in the "The following test(s) failed" section:

https://qa.nest-initiative.org/view/msysGit/job/msysgit-mingwGitDevEnv-test/lastCompletedBuild/

As you can see, there are way more tests failing than the 1 test that the TAP plugin lists in "TAP Test Result (1 failure / ±0)" on the same page.