Description of problem:
Working with Networking Policy as networking plugin, it's not possible to use more than one ingress selector, neither with OR / AND.

Version-Release number of selected component (if applicable):
OCP 3.11.51 / 3.11.104

How reproducible:
Always, with random behavior

Steps to Reproduce:
1. create two level structure : (1) APP, (2) DB
2. create networking policy for DB layer

   Using OR selector
   ...
   spec:
    ingress:
    - from:
      - podSelector:
          matchLabels:
            zone: DB
      - podSelector:
          matchLabels:
            zone: App
    podSelector:
      matchLabels:
        zone: DB
   ...

   or using AND selector
   ...
   spec:
    ingress:
    - from:
        - namespaceSelector: {}
          podSelector:
            matchLabels:
              zone: App
   podSelector:
     matchLabels:
       app: DB

 
3. Make a connection from layer (1) App to layer (2) DB

Actual results:
It shows a bit random behavior i.e. Middleware is able to connect to Database sometimes but if we re-deploy the Middleware deployment, it cant connect to Database at first and after some time like 4-5 restarts of container, it is able to connect to Database. 
This is a total random behavior, i.e. sometimes gets connected at once and sometimes takes 5 mins (4-5 restarts), sometimes 10 mins (7-8 restarts).

Expected results:
Connect as usual, at once.

Additional info:
* used different cloud providers: AWS / AZURE / On-Prem.
* used different DB servers: MySQL / Hazelcast