The design is for Harvest to try forever. But, there are some other modules it uses (SSL, NetApp SDK to name a few) that may consider some situations fatal. If I knew the place it's failing I could potentially wrap this to prevent it but I'm inclined to look for a solution outside of Harvest.
One solution could be to use supervisord to [re]start each of your harvest pollers with a config variable of autorestart=1 (here and here).
Another solution that is simpler if you are OK with missing soa few minutes of data is to just add a cron entry to run "/opt/netapp-harvest/netapp-manager -start" every 10 minutes. This script just parses the netapp-harvest.conf file, runs ps, and then starts pollers that are not already running.
I know these are workarounds but I think they are the best options for you. Hope it helps!
Cheers, Chris Madden
Solution Architect - 3rd Platform - Systems Engineering NetApp EMEA (and author of Harvest)