Ted, the idea is to use air only when there is no other way. And another reason is for the entire scope of pressure vessel testing, no one is likely to have air at the pressures needed. Of course, there are ASME, ASTM, etc, standards for all types of testing, and they specify the safety margin. Some specifications require testing at 150% of the rated working pressure, sometimes more, sometimes less, but rarely less. The least I can remember seeing is 146%. Why 146% I don't know, but that is what the spec was. It all depends on the purpose of the vessel. Think about all the differences in applications. It might be a fuel tank on your car.....low pressure and relatively low risk. So no need to build a 3000 psig vessel, but it still needs to be leak tested. On the other hand, look at the high pressure gases used in welding. Nitrogen and oxygen tanks often see 2500 psig. And that high of pressure can be dangerous, should there be a failure. Or if the contents are toxic, radioactive, etc, a whole new world of risk assessment opens up. And naturally, there are specifications governing each application.
But I an getting away from you question. Liquids are generally thought of as non-compressible, and any energy that is used to pressure up a vessel under test quickly dissipates should there be a catastrophic failure. In fact, it's rather unimpressive to watch. So, the idea is to fill up the vessel as much as possible with a liquid with water being most common because it is cheap and relatively harmless. There are a few exceptions with vessels where water cannot be used as they cannot be satisfactorily dried after the test. Other liquids are chosen when needed. When the vessel is as full as possible (no airspace is desired) then the system used to pressure the vessel is plumbed in. In the JPFab method, he uses a Porta-Power hydraulic hand pump (they also make powered systems, quite useful in their own right), and after topping off all possible remaining airspace and hoses, fittings, etc. with oil, he will pump it up just like extending the hydraulic cylinder normally attached to the pump. In theory, with zero air, all the pressure is contained in a closed system. In practice, it's just about impossible to get all the air out, so there is actually a small amount of air at system pressure. But hopefully, the energy contained in such a small volume is not enough to do any damage should there be a failure. Just a little squirt of oil or water. The expansion rate of air is a known factor, but depends on a few variables that are inconsequential for this discussion. That's part of how much force is available in a given volume of air at a given temperature, and how much force that could potentially make some fragment of a tank fly at you. With fluids, the pressure dissipates virtually instantly, so nothing much is going to come flying off, compared to air. The pressure achievable with hydraulics is quite high, quite cheap, and quite efficient, making it a good means of obtaining test pressure. On the other hand, air or generally any other gases pose problems when being compressed to the extent needed to adequately carry out tests. And the danger factor is immense compared to liquids. It isn't practical to test large vessels with gases anyway. Not all testing is about rupture/failure also. Some are simply leak tests, more like the rightful concern with the gasoline tank on our cars and trucks. I don't need my fuel tank to hold up under 10,000 psig. But I don't want a single drop to leak out.
There's much more to this than I have covered, but this is a PotD thread, so I'll stop rambling.