Agree - if you want to install the utility once and thereafter run it unedited then you get free upgrades essentially afterward if the copy that comes with weewx changes. Same for the systemd or init.d startup files or anything else that needs to hook into the os as either a mandatory thing or optional user-installed utility.
Of course. If you choose to edit it you are taking the responsibility to own your edited copy and handle any breakage if weewx moves under your feet, so to speak, in a future release.
I've been a sysadmin for 30+ years and would say there are a variety of appropriate/acceptable approaches especially when logging is involved.
Personally I like componentization and the ability to turn stuff on+off as well as the ability to add new things. That's why I like .d directories so much (or sites-enabled if you think webservers etc.).
The risk I think I'm trying to bring up is there seems to be a natural tension between making weewx infinitely extensible+configurable vs. keeping it 'wee'. My opinion is this particular one's original request violates the 'keep it wee' mindset perhaps, maybe, depending on implementation.
It's kinda like crontabs. You violate 'wee' if you just edit the system crontab. You are nice and componentized if you use crontabs (directory) and drop in pieces as components you can easily add/supersede/delete into what happens as an aggregated whole.
If logwatch could be made to work that way, seems worth thinking about trying to do when getroundtoit happens.