Downtime of almost 30min of pytorch and tensor flow builds during catalog source upgrade

Description of problem:

During the RHODS upgrade from v1.3.0-6 to v1.4.0-2 we observed that builds are getting recreated during the upgrade and it take almost 30mins for pytorch and tensor flow builds to up and running

Note:

1. In this time user can not Launch the pytorch and tensor flow notebook from JH (select buttons are disabled)
2. If notebooks (pytoch/tensorflow) are already launch prior to upgrade it will work without any problem

Prerequisites (if any, like setup, operators/versions):

Steps to Reproduce

1. Perform RHODS upgrade

Actual results:

During the upgrade user can not launch pytorch and tensorflow notebook

Expected results:

User should able to Launch all the notebooks images during the upgrade as per the choice. There should not be any disruption in the service 

Reproducibility (Always/Intermittent/Only Once):

Always

Build Details:

Workaround:

Additional info: