Richard Feynman stressed that we shouldn't make preconditions about how our future description of Nature is going to look like:
Lisa Randall, a top phenomenologist whom I know very well, gave an interview to Nude Socialist in which she says that it's an illusion that physics is mostly about the search for the final theory (among other things: read it). To a large extent, her answers are similar to Feynman's.
Phil Gibbs wrote that we need to find a TOE, after all.
Despite the contradictions in the "spirit" of their answers, I agree with all these three folks but I still think that some of the uncertainty in the first two people's comments are, to a certain extent, obsolete.
So, I agree with Feynman that it can't a priori be clear whether the Universe obeys the laws of a concise final theory that may be found after a finite time. It's critically important in science not to confine your reasoning by some assumptions whose validity isn't really certain – and sometimes not even justified – i.e. by dogmas.
On the other hand, I think that the evidence has accumulated that the alternative non-TOE scenario of the onion with infinitely many layers can't operate in Nature. Ken Wilson taught us to organize the phenomena in Nature according to their characteristic distance scale (or time scale or energy scale).
We may seemingly go to ever shorter distances and discover new and previously unknown layers of the onion, matryoshkas inside the larger matryoshkas, and so on. However, I am confident that we pretty much know that this "seemingly infinite" process inevitably stops at some point – the Planck scale. There are no distances shorter than the Planck scale that may be physically resolved, that make sense in the usual physical sense.
So once you describe all the effective theories – layers of the onion of knowledge – for all distance scales up to (longer than or equal to) the Planck scale, that's the end of the story. The last layer – the explanation how Nature behaves at the scale of the quantum gravity – will be the only task that you have to solve fully. No additional infinite hierarchy of effective theories can be squeezed over there. Because of the uncertainty principle's impact on the proper length that becomes severe and of order 100% near the Planck scale, energies that exceed the Planck energy no longer allow you to probe shorter distance scales and qualitatively new physical phenomena; instead, if the center-of-mass energy of a collision dramatically exceeds the Planck scale, you create a black hole (an ever larger one if you increase the energy) whose rough behavior is captured by the low-energy effective equations once again. No new physics emerges.
These were general comments boiling down to the Renormalization Group and the existence of gravity in our quantum world. But we have been collecting some precious, much more specific evidence for a few decades when it looked increasingly indisputable that string/M-theory is the final theory of everything. It apparently possesses everything that a final theory needs to have. It seems to be 100% robust and predictable: there's no way to "deform it" without spoiling its consistency. It allows no adjustable yet non-dynamical continuous dimensionless parameters. It seems that everything that's left is to understand string/M-theory more completely (including persistently confusing aspects such as the initial conditions and Big Bang singularity and the vacuum selection mechanisms in general if there are any) and isolate the solution that is relevant for the environment seen in Nature around us.
I could spend lots of time with mostly linguistic disclaimers that I don't find too deep or interesting here.
Of course, the term "theory of everything" has to be interpreted correctly – we/physicists don't know "everything", just the elementary laws to which (plus the knowledge of the initial state and "the right questions") everything may be reduced in principle. I think that people realize that a TOE has to be interpreted this carefully (as the theory about the elementary forces and building blocks – or the maths replacing them – only) and science doesn't completely stop when you find a TOE (lots of complex questions always remain) which is why I think that the opposition to the term TOE because of this reason is really unsubstantiated. Also, it's obvious that even in particle physics, many physicists are working on many things that have almost nothing to do with the search for a final theory – and that don't even depend on its existence in any way.
Still, the fact that physicists are working on various things doesn't mean that they're equally important. I agree with Phil that the search for a TOE is a very important research industry in physics, one that – according to the present evidence – should be solvable in principle but one that is so ambitious that it's clearly impossible to promise any deadlines for the date when the problem will be fully solved. We're not there yet but the "TOE research program" has already generated lots of profound spin-offs that have been valuable even for those who don't give a damn about a TOE.
I have always found the possibility that there is a TOE important enough for much of my thinking time to be occupied by questions related to this project – although the particular things one is thinking about are always much more concrete, limited, and modest than overarching claims about a TOE (a fact that the laymen may easily misunderstand, too: physicists are simply not meditating about a TOE, OM, TOE for hours in their office, they're looking at well-defined, seemingly more special, questions). On the other hand, Lisa Randall isn't passionate about a TOE – a fact that is correlated with her being a phenomenologist rather than a string theorist. Needless to say, I view both approaches as important ones but the TOE-focused, string-theoretical approach to be significantly more successful as a generator of progress in the last 35 years or so, a trend that is more likely than not to continue in the coming years, I think.