Description of problem:

Several e2e-olm tests kept failing due to the catalog pod being unreachable in https://github.com/openshift/operator-framework-olm/pull/108 

e2e test logs: 

```
waiting for catalog pod mock-ocs-main-preexistingcrdownerisreplaced-645dr to be available (for sync) - CONNECTING
waiting for catalog pod mock-ocs-main-preexistingcrdownerisreplaced-645dr to be available (for sync) - CONNECTING
waiting for catalog pod mock-ocs-main-preexistingcrdownerisreplaced-645dr to be available (for sync) - CONNECTING
waiting for catalog pod mock-ocs-main-preexistingcrdownerisreplaced-645dr to be available (for sync) - CONNECTING
.
.
.
```

There's a good amount of the following error in the catalog pod logs: 

```
time="2021-06-30T21:56:51Z" level=error msg="failed to list bundles: rpc error: code = Unavailable desc = connection error: desc = \"transport: Error while dialing dial tcp 172.30.108.164:50051: connect: no route to host\"" catalog="{mock-ocs-main-grlrk openshift-operators}"
```

This may indicate that registry-service was not accepting traffic/iptables needed to be flushed. 

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


How reproducible:

Not reproducible locally using kind cluster. Reproducible easily on 4.8 CI 

Steps to Reproduce:
1. Run tests on ocp 4.8 cluster
2. Run tests on a PR
3.

Actual results:


Expected results:


Additional info:

1. main branch (4.9 CI) did not have the same issue, but is worth making sure the problem does not exists in the main branch.
2. Test runs history: https://prow.ci.openshift.org/pr-history/?org=openshift&repo=operator-framework-olm&pr=108