original submission:

> (There are other problems with footnotes in multi-column mode; as
> can be seen above, the bottom of the column retreats up the page.

...

> These issues are distinct and should be filed as separate tickets.)

I was going to do just that, but before/after comparisons show that this particular issue was fixed as a side effect of the commit cited in comment #1.  It's not clear how, based on the code changes of this commit, which appear to have no effect on the root cause of this as explained over in bug #58736: "when a trap is set in multi-column mode for placement of the footnote, that trap is not cleared after the footnote is rendered.  The result is that the bottoms of your columns creep up the page as you go from left to right."  And the commit log talks only about the width of the footnote separator, so the fix to the column height might be a surprise to the commit's author as well.  But I won't argue with success.

> And on typesetters, the baselines of the columns don't
> align--there's a sliver of extra space at the top of the second.

I haven't looked into this aspect yet, so not sure if this still needs a bug report opened, or if the same commit also covertly resolved this as well.

This one is pretty simple.  The `$s` macro draws a horizontal rule of 2i without considering the possibility that the column width is shorter than that.

Interestingly, what I thought should have been a minimal reproducer didn't work; apparently, I need the footnote separator to be drawn in a column after the first to rouse the formatter's anger.

Input:

$ cat EXPERIMENTS/separatrix2.me
.if n .ll 65n
.de XX
.if \\$1 \{\
.  nr N \\$1
.  nr L \\$1
.  af L A
.  nop Jackdaws love my big sphinx of quartz\\nL.\c
.  ie \\nN%11=0 \{\\**
.    (f
.    nop \\** Sometimes.
.    )f
.  \}
.el \" newline
.nr N -1
.XX \\nN
.\}
..
.2c 4n 3
.pp
.if n .XX 31
.if t .XX 43

groff 1.22.4 output:

$ nroff -me EXPERIMENTS/separatrix2.me |cat -s
troff: EXPERIMENTS/separatrix2.me:20: warning [p 1, 1.2i, div '|f', 0.0i]: can't break line

     Jackdaws  love    sphinx of quartzK.2
my  big  sphinx  of    Jackdaws   love  my
quartzAE.  Jackdaws    big    sphinx    of
love my big  sphinx    quartzJ.   Jackdaws
of quartzAD.  Jack‐    love my big  sphinx
daws  love  my  big    of  quartzI.  Jack‐
sphinx of quartzAC.    daws  love  my  big
Jackdaws  love   my    sphinx  of quartzH.
big    sphinx    of    Jackdaws  love   my
quartzAB.  Jackdaws    big    sphinx    of
love  my big sphinx    quartzG.   Jackdaws
of quartzAA.  Jack‐    love  my big sphinx
daws  love  my  big    of quartzF.   Jack‐
sphinx of  quartzZ.    daws  love  my  big
Jackdaws   love  my    sphinx of  quartzE.
big    sphinx    of    Jackdaws   love  my
quartzY.   Jackdaws    big    sphinx    of
love my big  sphinx    quartzD.   Jackdaws
of  quartzX.  Jack‐    love my big  sphinx
daws  love  my  big    of  quartzC.  Jack‐
sphinx  of quartzW.    daws  love  my  big
Jackdaws  love   my    sphinx  of quartzB.
big    sphinx    of    Jackdaws  love   my
quartzV.1  Jackdaws    big    sphinx    of
love  my big sphinx    quartzA.
of quartzU.   Jack‐
daws  love  my  big
sphinx of  quartzT.
Jackdaws   love  my
big    sphinx    of
quartzS.   Jackdaws
love my big  sphinx
of  quartzR.  Jack‐
daws  love  my  big
sphinx  of quartzQ.
Jackdaws  love   my
big    sphinx    of
quartzP.   Jackdaws
love  my big sphinx
of quartzO.   Jack‐
daws  love  my  big
sphinx of  quartzN.
Jackdaws   love  my
big    sphinx    of
quartzM.   Jackdaws
love my big  sphinx
of  quartzL.  Jack‐    ____________________
daws  love  my  big       2 Sometimes.

____________________
   1 Sometimes.

(There are other problems with footnotes in multi-column mode; as can be seen above, the bottom of the column retreats up the page.  If you increase the argument to the `XX` macro, you can observe this develop progressively across all three columns.  And on typesetters, the baselines of the columns don't align--there's a sliver of extra space at the top of the second.  These issues are distinct and should be filed as separate tickets.)

The fix for this one appears straightforward.

@@ -1661,7 +1668,10 @@
 .      nr fp 8                 \" footnote pointsize
 .
 .de $s                 \" $$$ footnote separator
-\l'2i'
+.nr __ 2i
+.if \\n($lu<\\n(__u .nr __ \\n($lu
+\l'\\n(__u'
+.rr __
 ..
 .
 .