One thing to remember about metal parts: if someone really wants to rip something out, chain slings (a static material) tend not to resist shock loading very well. I'm a climber and it's incredibly easy to generate extreme shock loads. If you want some idea of how serious I'm talking, check out this video:
http://www.dmmclimbing.com/video.asp?id=5.
Those slings absorb some of the force themselves, reducing the amount ultimately recorded. The more something is able to stretch and transfer the force to something else, the less likely you'll break that component. However, you have to balance that with the risk of the material being abraded (as stretchier materials are often those that are easier to cut). Climbing ropes are sold as either static line (not suitable for catching a fall, only for raising/lowering systems) or dynamic line (for catching a fall).
Climbing equipment is generally rated in terms of the number of falls it survives. Rigging equipment often warns that you should be extremely careful about shock loads--because slings tend to fail. If you anchor something to the floor securely and you're worrying about anything more than a bolt-cutter, you've just given them a great anchor point with which to shock load a sling (and not damage the object it's securing).
Granted, most bike thieves aren't that interested--I'd be really surprised to find one that was. I'm talking about massive overkill situations. But if I was really worried about it, I'd say a really big weight isn't the most absurd idea I've heard. (A 5 gallon bucket of concrete doesn't count.)
A freestanding weight would make it difficult to apply a significant shock load without damaging an item of value (i.e., you need to supply multiple anchor points instead of only one) and if it's heavy enough, you'll need machinery (or a lot of people) to carry it off without attracting attention. That leaves cutting (either via abrasion or heating) as the primary means of attack on the sling. Additionally, that might shift the target to the locking mechanism.
