How are theories in physics validated against existing knowledge in practice? Often you hear people mention how most attempts at Unified Theories haven't "accounted for all the experimental observations".
The question is, how do you know you have accounted for all of the experimental observations? Is there something along the lines of a master list of all the properties, states, events, relations, functions, etc. currently observed experimentally between all of the entities described in physics?
Or is the process of validating a theory against all previous experimental observations more the sort of thing that occurs over many years, as some individuals test the theory out against this observation, while others test it out against that one? More of an ad hoc process.
Is there a point at which the theory is "completely consistent with all known experimental observations"?
It seems like validating a theory after the fact would only be possible if there was some sort of master list. It also seems that having a master list to look at before and during the process of constructing a theory would be very helpful.
Update
What I am particularly interested in hearing about is how specifically, in practice, you typically go about measuring a theory up to experimental observations. Do you just have a bunch of papers on your desk you compare your ideas too, or is it all just word of mouth, or is there some workflow you get into? Not "in an ideal world" type thing, which I get (scientific method, theories are constantly evolving/adjusted, etc.). More, what is the workflow like in practice?
 A: We do not validate theories. We fail to falsify them.
The whole scientific method is about the crucial fact that we will never have all experimental facts, and that tomorrow can always yield an observation that will invalidate what we have held to be true for centuries. And so science becomes an endless process of thinking up hypotheses, statements that explain some observations and also predict new ones. And then we go and hunt those predictions, be it by repeating old experiments, be it by devising new ones. And every hypothesis whose predictions turn out to be demonstrably false is thrown out. 
This goes on endlessly. And over time, the hypotheses get fewer and fewer, and those that remain get refined, and may turn out to be equivalent after all, or to predict very different things for experiments yet to come. All hypotheses that have not been falsified (but correctly predict everything else) are equally true. We tend, by Occam's razor, to believe those that are "minimal" in the sense of number and complexity of assumptions, but there's really no telling between such hypotheses.
The hypotheses that have survived long enough - and explained/predicted enough to be comprehensive - are called theories. These may not explain everything, but in the area in which they are applicable, they are generally believed to be true in the sense that nobody expects them to be falsified in the future. (Again, this is not certainty. But often, good enough)
And the more we know, the more difficult it becomes to construct new hypotheses, since the set of experimental fact we already know is so large that there are quite many predictions of new hypotheses that we can already test without new experimental effort. So, sometimes, hypotheses only attempt to explain a subset of our observations, and explicitly make no predictions for the rest. And that's fine. That's how subfields of physics arise.
The "unified theories" you are talking about are not expected to predict every observation. Grand Unified Theories are "only" expected to explain electromagnetic, weak and strong force phenomena by merging them into a single, unbroken force at high energies. Only when we are talking about a Theory of Everything it might be said that we expect it to predict every observation. And even then, tomorrow could bring a new fact, something never before seen, so even a theory encompassing every observation made so far is not validated. It is "merely" not false.

The above is an ideal of how science should work. There are no personal interests, no accidents, no miscommunications and no human failures in there. In reality, there is no central repertoire, no "master list" of everything known so far, and it would be so vast and incomprehensible that it would be useless to most. There is a reason we specialize in subfields - we know so much already that it seems impossible for one human being to hold all that knowledge within their mind.
In reality, there is no other process for testing new hypotheses than to publish them and wait what others say about it. Sorting the good from the bad, the false from the not-yet-false is a giant collaborative effort that is never finished. But, since, even if we were able to perfectly test hypotheses by a "master list" without biases, we would not get something perfectly true, this is good enough. It is all we should expect, and it is all we get.
A: New theories don't just appear fully formed from the febrile brains of the more lunatic theorists. To convert an idea for a new approach into a construction capable of making predictions is a vast amount of work often taking years. Theorists are only willing to put in such a large amount of effort if there is a known target to aim at. For example this might be an experimental result that is known not to match current theories (e.g. what dark matter is), or some aspect in current theories that is unexplained (e.g. the mass hierarchy).
The point is that when a new theory is constructed the theorists aren't aiming to explain everything, but rather they're attempting to explain a few specific things. A new theory doesn't have to be perfect, it just has to be better than the current theory. In other words it has to correctly predict everything the current theory does plus some extra things.
In fact there is probably no such thing as a perfect theory. That's because all theories start by making simplifying approximations and that means there will inevitably be circumstances under which the theory breaks down. A good theory will predict its own limitations. For example general relativity predicts singularities that we believe are unphysical, and the standard model predicts its own breakdown at energies near the Planck scale.
You say:

most attempts at Unified Theories haven't "accounted for all the experimental observations".

but the problem is that no unified theory has accounted for any experimental observation that isn't already accounted for.
A: As far as data of particle physics goes the theory going by the name of Standard Model embeds the overwhelming  majority of data from experiments and observations (astrophysics). It is validated by the data, and its predictions pan out, the recent one the discovery of the Higgs boson.
Of course there are many parameters in the SM fitted from previous data. BUT it also has an explicit group structure, SU(3)xSU(2)xU(1), which is crucial so that data can be described by the theory.
Thus what  any proposed new theory should do to show compatibility with most data is show that the group structure of the SM can be embedded , and the parameters of the new model can be adjusted so that the crossections and decay widths etc as they come from the SM can be reproduced by the new theory, which, in addition makes new measurable predictions outside the range of the SM. It would be good if it would solve the CP violation puzzle which is not predicted correctly by the SM, for example.
One should not also forget that Lorenz invariance should be included  in the model in the region where special relativity holds, for a proposed theoretical model to be viable against the data.
The reason why many believe that a string theoretical model will be the model for a theory of everything is because it does fulfill these conditions, except it has so many parameters/possibilities that it has not been nailed down to a single model up to now.
Edit after update of question
I am an experimental particle physicist, now retired. This means I have seen the evolution of the standard model starting from the very early times. It was an interactive process between experiment and theory. Experiments showed symmetries that theorists found could be described with group structures.
For example:

In physics, the Eightfold Way is a term coined by American physicist Murray Gell-Mann for a theory organizing subatomic baryons and mesons into octets (alluding to the Noble Eightfold Path of Buddhism). The theory was independently proposed by Israeli physicist Yuval Ne'eman and led to the subsequent development of the quark model.
In addition to organizing the mesons and spin-1/2 baryons into an octet, the principles of the Eightfold Way also applied to the spin-3/2 baryons, forming a decuplet. However, one of the particles of this decuplet had never been previously observed. Gell-Mann called this particle the Ω− and predicted in 1962 that it would have a strangeness −3, electric charge −1 and a mass near 1,680 MeV/c2. In 1964, a particle closely matching these predictions was discovered1 by a particle accelerator group at Brookhaven. Gell-Mann received the 1969 Nobel Prize in Physics for his work on the theory of elementary particles.

A large number of experiments had already been performed and the result published before theorists managed to perceive the underlying symmetries and organize a model for the data, which was predictive .
So in elementary particle physics progress comes from  interaction, experiment and theory. Theory gives a predictive background for experiments to be performed, the experiments are carried out and validate or falsify the theory, while also the same experiments may measure some aspects which need new theoretical frames. And so on.
The LHC experiments are collaborations of 3000 physicists, working on the complex detectors and the programs necessary to analyze the data and check predictions. The Higgs was found, and it is a prediction of the standard model and in that sense it is validated ( the word used in contrast to proved correct, a physics theory can only be falsified by or be compatible with the data). Many theoretical physicists have contributed to the construction of the standard model theory, so on both sides, experiment and theory, it is a laborious interactive process that converges to specific models that fit the data, and then gives predictions for new measurements that might falsify the theory.
A: It is very hard to keep track of all the experiments that have been done so far and the way in which they were done and under what assumptions. 
For instance, Newton's inverse square law is a "weak field limit" of general relativity. It works really well on a lot of scales. But on which scales was it actually precision-tested? Also, the experiment is often designed to test only one aspect of the theory and fixes the rest in it's consideration. Say you want to test Newton's gravity. Do you want to test the qualitative relationship $\sim 1/r^2$ between two objects? Or do you want to test gravity acts the same on bodies of all masses? Of all materials? Or do you want to know if gravitational mass is the same as inertial mass and on which length and mass scales this was tested? Does gravity act on charge? Does it's influence change on moving bodies? 
It is simply impossible to remember whether the experiments really test all of this and any list may omit tests of another aspect you might be interested in. There are review articles of large classes of tests but these usually talk about tests of specific concurrent hypotheses. Obviously, you cannot discuss all possible hypotheses at once.
There tend to be well-known "motivational" lists made up of experiments every theorist checks before pursuing a theory further. These lists show unnatural results or even violations of the current paradigm. For example in cosmic inflation (and inflation-like) theories, you will check against the experimentally observed primordial fluctuations because this was your motivation for the formulation of your theory. In developing alternative gravitational theories, you check against experiments labeling an effect as "dark energy" or "dark matter". These "motivational" lists tend to be very well reviewed and discussed in literature.
But otherwise than that, to make the quest of "validating" (i.e. testing) a completely original and novel theory tangible, you actually have to find out where it differs from the predictions of the current paradigm, and sniff through the literature whether this case was tested. Once you have one single effect at a special scale, it is indeed possible to find or at least devise a certain "master list". You can be lucky as Einstein in the case of the precession of the perihelion of Mercury and find that the deviation predicted by your new theory perfectly fits the data the original theory did not account for. But sometimes the deviation is in untested ranges and you might need to propose a new doable experiment.
In conclusion, there is no "master list" - in most cases it would just be too long. A good start is to have a solid background in the current paradigm, it's flaws, and keep track of how experiments of all sorts actually work and what scales do they probe. Taking a few lab courses on the way is quite essential even for a pure theorist - if even just to understand how general experimental procedure works. Often it is ok to just have an overall idea of where the current paradigm works well, and the real showdown comes when you predict a deviation around these ranges. 

By the way, I have just stumbled upon software of some SuSy phenomenologists which does this "master check" against LHC data for new particle models - CheckMATE. Just by reading the linked article you can see how hard the objective of checking against data nowadays is.
