on rc.sysinit there is a posible "race condition" between calling getkey
(for interactive startup) and killing getkey (before starting selected
runlevel scripts)

a snip of the rc.sysinit is shown below

{
# Clean up utmp/wtmp
>/var/run/utmp
touch /var/log/wtmp
chgrp utmp /var/run/utmp /var/log/wtmp
...
kill -TERM `/sbin/pidof getkey` >/dev/null 2>&1
}&
if [ "$PROMPT" != "no" ]; then
  /sbin/getkey i && touch /var/run/confirm
fi
wait

as you can see, if the code between the {} runs too fast, you could end
with an stalled "getkey" waiting forever for input (as i have experienced
on some Compaq Dual Processor, Hardware RAID servers like Proliant DL360
and DL380).

after testing with bash1 and bash2 scripts there seems to be at least a
needed 1 second delay before the kill is reached for getting a successfull
kill of getkey (running on different threads of the same script)

getkey from initscripts-5.49 has a timeout parameter that could be used to
avoid this "race condition", so a patch would be needed to upgrade getkey
and to use the timeout parameter on rc.sysinit

both patches are attached and a "proof of concept" is available from :

ftp://sajino.terra.com.pe/pub/linux/redhat/carenas/6.2/initscripts-5.00-3.i386.rpm

thanks goes to Sandro Jurado from Compaq for pushing on the right direction
and away from a problem on their Compaq Smart Array Adapter ;)