RE: narrowing considered harmful

[Drew Adams]

> >> It can even be a bad idea (set-marker to nil requires scanning
> >> the whole list of markers, so it can take a while.  It can be
> >> more efficient to let the GC collect them later on).
> 
> > Good to know this. The manual contains this:
> 
> > ,----[ (info "(elisp)Overview of Markers") ]
> > |      ;; When you are finished using a marker, make it point nowhere.
> > |      (set-marker m1 nil)
> > |           => #<marker in no buffer>
> > `----
> 
> > Should that be changed?
> 
> Not necessarily: whether it's better to nil them explicitly or to leave
> them as they are depends on the specific case.

For which some guidance can be given (in the manual)?

Perhaps say something like what you said here, regarding performance
as one reason you might not want to "make it point nowhere"?

The text is currently imperative: "make it point nowhere".
What you are suggesting is something different.
Please consider providing some nuance (better guidance).