More robust APIs, too many intermittent errors creating/removing ResourceGroups
It's very difficult to automate against Azure due to random failures and inconsistencies.
For example tear down a Resource Group and wait for delete to complete. When Azure says it's delete run a script to re-create the ResourceGroup with same name. Even though it says deleted, the APIs complain the resources of same name are already in use.
After an indeterminable amount of time after delete, the same script can successfully recreate the ResourceGroup. This causes a lot of problems. If it reports deleted, then it should be deleted. State should always be known and accurate.
Likewise when creating resources it randomly fails. The same script run a second time succeeds.