That would be the main point there. With the other RAID formats, there would be zero downtime, where with RAID0 you are guaranteed downtime, even with a perfect backup.
We do agree for the most part here.
If one were to stripe internal drives or use a single drive, then have a good backup, they could boot from the backup if the internal drive(s) fail. The catch of course is the time between the drive(s) failure and the last back up and any data created in that span of time being lost.
As for using something like RAID 5 (as example), if a drive failed the RAID 5 remaining drives would keep working but often at a slower speed. Some RAID 5 set ups don't even really work until the drive is replaced (this is rare but I have seen this before).
Obviously, there are trade offs. If one had Tbolt set up, they could run striped drives internally and mirror to an equivalent counterpart externally without taking much of a hit on speed. If a drive fails, switch to the mirrored set of drives and continue. Again, there is still the chance for yet another drive to fail before replacement of original failed drive occurs.
There are many variations on this theme and everyone has a slightly different approach. I suppose that running all external, RAID 6 with fast mechanical drives or better yet - SSD's, would be very safe. Then again, there is always opportunity for the external enclosure or DAS to fail and it is just another iteration of realizing that every combination has a critical point.