I agree that -o pipefail is not great since it's totally fine for pipes to break in some cases (e.g. I've occasionally seen it cause problems with pipelines like `grep | head`).

Hard disagree about -e though. I get it's not perfect but it has probably saved me dozens of hours of frustration and heartbreak. If you forget to handle errors for a particular line, then the script will just keep going and that's how you accidentally delete your entire home directory. Sure you can rely on shellcheck but for me personally I don't have shellcheck available 100% of the time as I often find myself writing scripts as raw strings (e.g. ssh -c).

I'm a fan of Bash3 Boilerplate (https://github.com/kvz/bash3boilerplate) and use a modified version of it for my shell scripts now. I like the code style and the logging, although I've amended the logging slightly to add a pipe input for long running processes that you want to see the output from before it finishes.

I prefer having a stop-on-unexpected script as it makes errors far more explicit and it's not too onerous to work round the peculiarities of return codes.

Greg's wiki (https://mywiki.wooledge.org/) is my go to resource for looking up snippets and learning to avoid the footguns - that and shellcheck are the key to "robust" bash scripts.