Description of problem: When doing a new installation on a RAID array anaconda writes a grub boot sector on _one_ of disks. The problem is that this quite often this is not that disk which later BIOS attempts to use as a boot disk. This is a pretty regular occurence. With some experience it is pretty clear what happened and how to fix things but in many real life situtations people were stumped about what to do with "non-working installations". It would be really desirable if anaconda would write boot sectors on _all_ disks which could end up as boot devices, or at least if there would be a way to request such thing. Especially in situations when RAID is used for mirroring it would be also really good to be able to boot from a mirror disk without necessity to go through extra hoops in order to make that feasible. By that time installation media may be not even around and not everybody is "grub-wise" enough to prepare for that beforehand; in particular when a new installation accidentally booted right away without any hiccups.
I believe this is working already in Fedora 9. I just did a F9 install on a RAID1 mirror and it looks like anaconda did install grub to the MBR of both members of /dev/md0. In my case grub got installed on /dev/sda and /dev/sdc with each installation of grub referring to its own device for stage2 and grub.conf. /boot is on /dev/md0 whose members are /dev/sda1 and /dev/sdc1.
Closing this bug. If the behavior presents itself again, please reopen.