Follows a list of cases covered by the ATH:

• Create a Freestyle job that connects to a remote git repo via ssh credentials and is able to checkout the code
• Create a Freestyle job that connects to a remote git repo via ssh credentials and is able to checkout a custom branch of the code
• Create a Freestyle job that connects to a remote git repo via ssh credentials and is setting a custom name for the remote
• Create a Freestyle job that connects to a remote git repo via ssh credentials and is able to checkout the code to a local dir
• Create a Freestyle job that connects to a remote git repo via ssh credentials and is able to poll for changes
• Create a Freestyle job that connects to a remote git repo via ssh credentials and is able to properly display the git revision on the status page
• Create A Freestyle job that is able to checkout submodules
• Is able to properly calculate changelogs
• Tests the clean after checkout functionality
• Is able to create tags for builds
• Is able to set up a custom scm name
• It tests sparse checkouts
• It tests Ancestry and Inverse chooser strategy

Unfortunately it seems the records of past ATH runs have been deleted so my review of how stable is the ATH regarding git is not going to be very accurate, I will create a custom run in my private instance but obviously, results are not going to be as trustable as having real data in ci.jenkins.io

As at this moment, there is not enough historical data in ci.jenkins.io I have created a job that runs the ATH GitPluginTest every hour to get an idea of how stable they are. I am going to let this running 24 hours and will update here with the results

My environment is having some problems, so I am using the new runATH step to launch some runs in ci.jenkins.io to get a very quick idea of flakiness even if we do not have a lot of historical data.

See JENKINS-50536 for the list of P1 cases and it's coverage

So it seems is not only on my environment, currently  all tests in GitPluginTest are failing in ci.jenkins.io since build 72

From the history data (it seems some data like the age is still there) I can say that GitPluginTest#create_tag_for_build has been failing for the last 15 builds, so no flakiness it seems but a real failure

Interesting enough it seems all git tests are failing due to the same error, see the linked result of build 72 

Created JENKINS-50789 to address the missing P1 cases