many build failures now are due to testEditor or github creation tests. 
Retries help a little - but there are still too many failures. 

Perhaps it is time to build some robustness into the api calls themselves that talk to github? (it isn't clear what the failures are...), more investigation needed.