The actual Gaia hypothesis is that life tends to change the environment such that the environment becomes more suitable for life in general.
More specifically, the sum of the actions of the biosphere is increasing the chance of the biosphere continuing to exist, by promoting conditions favourable to its existence. As opposed to the trivially true conclusion that particular organisms alter the environment to benefit that particular organism.
Of course the Gaia hypothesis is also trivially easy to prove in that there need to be plants before there can be herbivores, and there needs to be death before there can be decomposers. But, this just shows that life tends to diversify, and the more diverse it is the more of it is there to diversify, and of course more diversity increases the number of chances to resist any particular disaster.
The other actions usually attributed to Gaia are quite dubious, especially the one where Gaia is supposed to be altering geology directly to improve life's chances. There may life-caused negative feedbacks enforcing equilibrium, but how strong are they? Even if they are strong enough to dominate, how do you rule out mere chance, meaning that current life could crash at any time? Aside from this, the difficulties inherent in group selection make it unlikely that life can adapt to improve the survival of species as a whole, even by accident. Also, one should ask how useful this particular abstraction is. Even subunits of Gaia - like nations - don't necessarily act coherently enough to be useful to consider as a whole.
Personally I'm interested in this because my definition of life has at least one flaw; it includes Gaia. I define life as anything that can be assigned a goal, which it will defend. (All consciousness is therefore automatically life, as it actually has goals.) Another way to put it is that all life has needs or wants. Gaia can be assigned the goal of diversity. For instance, mass extinctions tend to increase speciation rates; Gaia reacts to extinction. I could say, "Well, it's a completely blind process, there's no entity that is actually served by this goal." Except life itself is also a dumb process, so I'd throw out the baby with the bathwater.
I just realized I at least need to add that for something to be a goal, it has to be at least in principle possible to fail. The universe may want to increase entropy, but it can't fail to increase entropy, and thus can't die, and thus can't be alive in the first place.
I think this may successfully eliminate Gaia. The only way to interrupt the process is to exterminate all life on Earth. You cannot have the constituents without getting Gaia automatically, whereas putting the molecular constituents in even slightly the wrong order in a cell results in a nonliving cell. By analogy, it would be like assigning a goal to a cell such that it could not fail the goal unless you annihilated every constituent molecule. It stops being a cell long before it stops being alive, which essentially means it cannot fail the goal, and thus it's not much of a goal.
On the other hand, the fact it is so close helps explain why people find the Gaia hypothesis so seductive. Almost all the features usually attributed to living things can also be found in the system as a whole. Almost.
Intriguingly, under this definition you can get immortal life, in the sense of being invulnerable, but only if it is also conscious. Otherwise you've just found a force of nature. I would be intrigued to find any other clearly silly situations under the definition, like finding ephemeral structures on a neutron start that can be assigned goals until they decay.