Description of problem:
Hi Team, CU is using hostAliases configuration in DC like below

hostAliases:
    - hostnames:
      - domaindnszones.xxx
      ip: 10.17.xxx.xx
    - hostnames:
      - ldapsg.domain.xxxx
      ip: 10.17.xxx.xx

Recently they have upgraded the cluster from 4.12 to 4.13 and they are getting below error while doing rollout:

 Reason:      Error
    Message:     error: couldn't assign source annotation to deployment ms-owcs-security-r2-20: failed to create manager for existing fields: failed to convert new object (platform-operations-bos/ms-owcs-security-r2-20; /v1, Kind=ReplicationController) to smd typed: .spec.template.spec.hostAliases: duplicate entries for key [ip="10.17.xxx.xx"]

Can someone please help me to know whether duplicate Ip for hostnames in hostAliases are supported or not? If not why it was working till 4.12? How can this be fixed in 4.13?
What exactly has changed in the recent 4.13 or upsteam which has caused this issue.
From the internal discussion over slack we found that, in upstream there are a few mentions of duplicate key check in PRs similar to : https://github.com/kubernetes/kubernetes/pull/121575
Engineering team  guess is at some point, in some bump in https://github.com/kubernetes-sigs/structured-merge-diff changed the behavior.

However we need a confirmation what exactly has changed it.

Version-Release number of selected component (if applicable):

4.13
    

How reproducible:

    Everytime

Steps to Reproduce:

    1. Use duplicate enteries in hostAliases for different host name.     
    

Actual results:

    Giving error as mentioned above

Expected results:

    It should work as expected as it was working in 4.12

Additional info:

    Discussion on slack is already on going , will attach the link in comment

Resolution: 

hostAliases:
    - hostnames:
      - domaindnszones.xxx
      - ldapsg.domain.xxxx
      ip: 10.17.xxx.xx