bin/57180: sh fails to foreground partially terminated pipeline job

[campbell+netbsd%mumble.net@localhost]

>Number:         57180
>Category:       bin
>Synopsis:       sh fails to foreground partially terminated pipeline job
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    bin-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Jan 10 01:45:01 +0000 2023
>Originator:     Taylor R Campbell
>Release:        current
>Organization:
The NetBg Fgndation
>Environment:
bomb cyclone aftereffects
>Description:
I started a shell job for `foo | less', more or less, with a long-running foo that didn't produce any output for a while.

I hit ^Z and tried to bg the job so it would continue running, but the whole pipeline stopped on less(1) tty output.

I manually ran kill -CONT on the foo process and it continued on its merry way for several hours.

Once foo terminated, I tried to fg the job, but it refused with:

fg: Cannot continue job (No such process)

The job is still listed in jobs, with foo as `Done' and less as `Stopped (tty output)'.  Using `fg %<jobnum>' fails the same way.

The full pipeline was a little more involved:

./binary16std | awk '{ x[$4]++ } END { for (i in x) print i, x[i] }' | sort -n | awk '{ print $1/(2**24), $2 }' | less

# jobs
[2] + Done                    ./binary16std |
      Running                 awk "{ x[\$4]++ } END { for (i in x) print i, x[i] }" |
      Stopped (tty output)    sort -n |
      Stopped (tty output)    awk "{ print \$1/(2**24), \$2 }" |
      Stopped (tty output)    less
>How-To-Repeat:
Haven't tried to reproduce but the description above should provide enough information to make a reproducer and track down the bug.
>Fix:
Yes please!