In general, you are right. Once you explain to the borrow check you are doing the right thing then you have a fair level of confidence in it being right.
I think the point from the article is if it is worth it. Give the single sentence in the article about a list and pointer. There are various ways to do it safely (and ways to do it unsafely like you pointed out) and sometimes it might be ok to put the onus of that onto the developer and validated with valgrind, asans, and unit tests. For some this is far more enjoyable than worry about the borrow collector.
It’s all a trade off and it’s ok to have difference criteria when approaching a problem like a cli tool or even a business critical service. The problem becomes when Rust people claim that the borrow checker is the only way and we all roll our eyes.
I used to work at FB a a while back (working on infra, not FE or anything user touching) and we would have private development servers and occasionally I would need to do work on the FE PHP repo (called www) and would need to stay a development server for FB.
It would connect ti the various services and database as read only and was a private copy. It functioned just like a real webswver with the exception that the cache was stone cold. It would takes MINUTES to load the front page the first time. It is crazy how much caching just makes it functional.
The problem is that there are so many services and datasets read all the time.