The discussion “User preference persists through Safe Mode” revolves around the idea that Discourse’s Safe Mode should bypass user preferences, such as customizations and overrides. emyoulation initially suggested this idea, citing that user preferences can often be the source of issues and that many safe modes fallback to defaults.
Falco disagreed, stating that Safe Mode is intended to remove plugins and themes, not user settings. However, emyoulation countered, suggesting that adding an option to disable loading end-user customizations in Safe Mode would be beneficial. Falco responded, offering to consider a pull request for this feature if it is well-implemented. Ultimately, emyoulation acknowledged that they are not qualified to code the feature themselves and hopes that someone else will take on the task.
HEY! Where can I find the feature that recognizes key phrases (like Safe Mode ) and automatically links them!?!?! I want that!!! 
The feature is called auto-replace watched words.
Close, but not quite! But I managed to find it.(Whilst dancing around because I was so excited that I was in danger of piddling myself!)
Linkify! [No, @Moin was right. The Linkify function was folded into Watched Words → Link,]
You don’t need the linkify component anymore, it’s been built-in:
admin → customize → watched words → link
Thanks! Could you admin-hack the posting linked above and insert a [Now built-in as “Watched Words”] notice at top? And add “Watched Words” to this site’s list of Watched Words? Amazing that the first thing linked wasn’t the feature itself!
No, it shouldn’t. The safe mode feature is about removing plugins and themes, and not user settings.
I’ve been a professionally beta testing software (from application suites to BIOS) for more than a quarter century. User preferences are often a sources user issues. Many safe modes fallback to defaults.
Since there are already 3 sets of options when entering Safe Mode. adding an option to disable loading end-user customizations seems like a no-brainer:
If that’s something you are keen to work on and contribute as a pull request we can consider merging it, provided the PR is in good shape.
Forking it myself would certainly be the most the immediate option… but probably also the most disastrous.
Let’s see if the idea piques the interest of someone at least mildly qualified to code it. (Which ain’t me.)
