What info is required for this "feature"? You can see more discussion in this thread https://lists.gnu.org/archive/html/bug-groff/2022-11/msg00008.html

You should not be so surprised!!

If you have a look at the file s.pdf attached to bug #62923, and my long comments in that bug you will see that I demonstrated the same thing. The equivalent glyph in font XITSMR does not behave this way, it is just a bar at the top of its own bounding box.

Spawned off of bug #62923.

Curious, I attempted equation construction by hand to see how I would fare compared to eqn(1).

I also commented out the redefinition of the "radicalex" special character in ps.tmac.

So check this out.

Input:

$ cat EXPERIMENTS/distance.ms
.LP
.nr w1 \w'\(sr'
.nr w2 \w'\[sqrt]'
A text radical is \n(w1 wide.
A special radical is \n(w2 wide.
.LP
A\(srB
.LP
A\[sqrt]B
.EQ L
d = sqrt { ( x sub 2 - x sub 1 ) sup 2 + ( y sub 2 - y sub 1 ) sup 2 }
.EN
\fId\fP = \(sr
.\" Define the radicand.
.ds rc (\fIx\fP\*<2\*> \- \fIx\fP\*<1\*>)\*{2\*})\
 + (\fIy\fP\*<2\*> \- \fIy\fP\*<1\*>)\*{2\*})
.nr ln \w'\*(rc'
\Z'\*(rc'\l'\n(lnu\[radicalex]'
.LP
C\(rnD (root en extender)
.LP
C\[radicalex]D (text radical extension)
.LP
C\[sqrtex]D (special radical extension)
$ ./build/test-groff -e -ms EXPERIMENTS/distance.ms >| distance.ps