It's happening again!  This time, I thought to run ktrace...

$?:T:$PWD
$ kdump ktrace.out
 17650 ktrace   RET   ktrace 0
 17650 ktrace   CALL  __sysctl(0xdfff7bc,0x2,0xb1a4,0xdfff7c4,0,0)
 17650 ktrace   RET   __sysctl 0
 17650 ktrace   CALL  break(0xb1dc)
 17650 ktrace   RET   break 0
 17650 ktrace   CALL  break(0xbffc)
 17650 ktrace   RET   break 0
 17650 ktrace   CALL  break(0xdffc)
 17650 ktrace   RET   break 0
 17650 ktrace   CALL  execve(0xdfff83c,0xdfffca4,0xdfffcb4)
 17650 ktrace   NAMI  "/home/seebs/bin/kill"
 17650 ktrace   RET   execve -1 errno 2 No such file or directory
 17650 ktrace   CALL  execve(0xdfff83c,0xdfffca4,0xdfffcb4)
 17650 ktrace   NAMI  "/usr/local/bin/kill"
 17650 ktrace   RET   execve -1 errno 2 No such file or directory
 17650 ktrace   CALL  execve(0xdfff83c,0xdfffca4,0xdfffcb4)
 17650 ktrace   NAMI  "/bin/kill"
 17650 kill     EMUL  "netbsd"
 17650 kill     RET   execve 0
 17650 kill     CALL  kill(0x4477,0x9)
 17650 kill     RET   kill 0
 17650 kill     CALL  exit(0)
$?:T:$PWD
$ ps -uxwwtp4
USER       PID %CPU %MEM   VSZ  RSS TT  STAT STARTED       TIME COMMAND
root     17527 48.7  1.9   344  280 p4  R     2:54PM   13:12.99 tar cf - src
root     17528 49.1  2.9   440  432 p4  R     2:54PM   19:05.52 gzip -1
seebs     4836  0.0  4.2   296  624 p4  Is    7:22AM    0:01.96 -ksh (ksh)
root     17587  0.0  2.7   456  400 p4  S     3:13PM    0:01.29 sh
root     17653  0.0  1.6   288  232 p4  R+    3:38PM    0:00.00 ps -uxwwtp4
$?:T:$PWD
$ ps -lxwwtp4
  UID   PID  PPID CPU PRI NI   VSZ  RSS WCHAN  STAT TT       TIME COMMAND
 6667  4836  4835   1  18  0   296  624 pause  Is   p4    0:01.96 -ksh (ksh)
    0 17527     1 196  77  0   344  280 -      R    p4   13:14.58 tar cf - src
    0 17528     1 203  78  0   440  432 -      R    p4   19:07.29 gzip -1
    0 17587  4836   1  10  0   456  400 wait   S    p4    0:01.31 sh
    0 17654 17587   3  28  0   288  232 -      R+   p4    0:00.03 ps -lxwwtp4
$?:T:$PWD
$

---
So far as I can tell, tar just plain ignores the kill signal, as does gzip.
Since neither is typically blocking SIGKILL, I assume this is a bug.

-s
(-current, Amiga, about 2 weeks old, compiled with -m68030 but running
on an '040.)