Mark this as TestBlocker, as currently automated cases can not be run on Azure due to this bug, and there is no regression test coverage on Azure for egressIP
Just to reiterate: this clearly looks like an issue in the Azure layer as this can easily be reproduced by just spawning 2 VMs, assigning a static secondary IP to the first VM from inside linux, and then attaching an ipconfig with that static IP to the first VM from Azure; one can observe significant delay between the success manage from Azure and traffic actually passing when one runs a continuous ping sourced from the secondary IP of VM1 to VM2.
I am taking the TestBlocker off of this bug. It's clearly a problem with the cloud platform and not on our end. I'll have to get back to this, but for the time being I'd suggest adding a sleep or something to the tests between the configuration and the validation step.