In reply to ogreville:
I think the issue is at least a little bit more subtle than the arguments sometimes suggest. Here are some considerations.
1. Is the backup actually going to work? I see a certain number tied loosely enough to make me wonder whether they will really grab in an emergency. But more treacherous is the method of putting the autoblock on a leg loop and the device on the harness belay loop. There are two possible problems with this.
(a) If the leg loop is a buckle leg loop, it is easy to put the autoblock on in a way that will pull open the buckle, opening the leg loop and disconnecting the autoblock sling.
(b) More insidious is the spacing between the autoblock and the rap device when the autoblock has locked. Typically, there is very little space. In such cases, raising the leg whose leg loop is the autoblock attachment will release the autoblock. In almost all cases, if the rappeller inverts, an autoblock set up on the leg loop will collide with the device and release.
If you are doing any of these things, you are far from reliably protected, but are likely to act as if your backup functional. There has been at least one fatality in the US because of this.
The well-known but not universally-adopted solution is to extend the rap device to chest or head level and put the autoblock on the harness belay loop. Assuming the knot isn't too loose, this method is far more robust.
2. Okay, you have a backup system that will actually work if you suddenly and unexpectedly let go. There are various ways this might happen. Some of them are the result of inexperience and/or improper training, paradoxically reinforced by the notion that it is ok to drop the brake strand because the backup is there. But it is also possible that a falling rock will deliver just the right blow to make you let go. I haven't heard of this, but it is conceivable. The most likely reason for dropping the brake strand, at least for an experienced climber, is loss of consciousness. Assuming this happens, you now have an unconscious rappeller fixed to the ropes midway down the rappel. If they don't regain consciousness very quickly and continue their descent, the party is faced with a complicated, time-consuming, and possibly dangerous rescue scenario, and the unconscious rappeller is in danger from suspension trauma.
3. When all these considerations are taken into account, I think the most rational policy is for the first person down the rappel to use a properly-configured autoblock backup, but for everyone after the first down to omit the backup and instead be protected from below with what the US climbers call a "fireman's belay." In case this term is not common in the UK, it refers to someone at the base of the rappel standing ready to apply traction to the rappel ropes in order to control the descent of the rappeller, independent of any action on the part of the rappeller.
In this system, the party still has to contend with the rescue of an unconscious rappeller for the first person down, but not after that. And in the case of the first person down, if circumstances actually make it seem as if they might be knocked out, it is arguably more sensible to lower them or have them rappel single-strand with a belay from the other strand rather than risk a situation in which they have to be rescued while hanging unconscious in mid-rappel.
The result of all this is a system in which not everyone uses a rappel backup but everyone is in fact backed up, while mitigating some of the most serious downsides to the more usual approach of autoblocks for all. A problem is that the system makes rappelling a team endeavor, and most people conceptualize it as an individual process.
4. Some people like an autoblock backup as a convenience more than as a safety measure. The autoblock allows the brake hand to be released to deal with tangles, take photos, and swat midges. I think this is a bad idea for two reasons.
(a) The leg-wrap is fast and more secure.
(b) Using backups this way undermines what should be a total commitment to never dropping the brake strand. We then hear of people dropping the brake strand and being "saved" by the backup, when in fact their training should have prevented them from ever releasing the brake hand as long as they were conscious.
5. This leaves people doing extensive amounts of hanging on rappel, say for equipping a sport route. In that case, specialized devices are a far better approach than an ATC and a prusik.
6. It might be worth ending by mentioning the myth that a rappel backup might prevent a climber from rappelling of the ends of the ropes (if there are no knots). In order for the backup to work, it has to be released. Tests have shown that people cannot do this, even under test conditions when they are belayed and know they are going to go off the ends of the ropes.