cf-nginx-ip-passthrough script fails randomly

I have a scheduled task which fetches and runs the cf-nginx-ip-passthrough (cf.sh) script periodically. It seems to fail at random, disabling the cloudflare.conf file and touching cf-stop. I believe this is caused by an occasional network hiccup but I can't be sure, and it highlights three problems:

1. The script always returns 0 so the Plesk setting to notify on error doesn't work.
2. If I enable notifications for every run, the emails only contain the cURL output and Configuration applied. Restarting Nginx. On and after the fatal failure, I only receive the cURL output in the emails, so I can't be certain what going wrong.
3. Oddly, the .disabled file is always valid when I log in to intervene (I rename it back to cloudflare.conf and nginx -t succeeds), so I wonder if the code to rollback to the previous working config is broken.

Start with the cf.sh script and reproduce its scheduled execution while inspecting the cURL output, cloudflare.conf, and cf-stop behavior. Check the rollback path with nginx -t and determine what happens during an intermittent network failure. Done means failures return a nonzero status, notifications include the relevant error, and a valid previous configuration is preserved.