Can evolution go backwards?

A big question in evolutionary biology for some time has been whether or not a gene can revert to an ancestral state if the (environmental) pressures that selected for the initial changes to the gene were removed and the ancestral condition (or gene function) was again favored.  It now appears, that, at the genetic level, things can't go back to the way they were... at least not in the case of the gene that codes for the glucocorticoid receptor (glucocorticoids, a.k.a "stress hormones", got a brief mention here a few days back).
In a new article in the journal Nature researchers have shown how the glucocorticoid receptor evolved from a protein that was likely responsive to another hormone known as aldosterone.  The group identified 7 key mutations that were responsible for losing this sensitivity to aldosterone, but when they artificially manipulated the genetic sequence to reverse these mutations, what they got was a non-functional gene (coding for a non-functioning protein).  In order to figure out why the reverted gene didn't have the predicted function, the team, headed by Joseph Thornton of the University of Oregon, further analyzed the sequence and discovered 5 other mutations that were not directly involved in the function of the glucocorticoid receptor, but were apparently important enough to prevent a reversion to the ancestral aldosterone/glucocorticoid sensitive protein.  You can find a more detailed synopsis at Science Daily, from which I will quote:

"Suppose you're redecorating your bedroom -- first you move the bed, then you put the dresser where the bed used to be," Thornton said. "If you decide you want to move the bed back, you can't do it unless you get that dresser out of the way first. The restrictive mutations in the GR prevented evolutionary reversal in the same way."

So does this mean it is impossible for evolution to go backwards?  (to evolve the same gene or protein that was lost?).  Well, at the very least, it is so improbable (on the order of billions if not trillions to one odds) that it is, basically, though, I guess, not ultimately impossible.  The reality is that, if an ancestral condition is once again favored, it is much more likely that evolution will find some other way for the species to get what it needs to survive and reproduce... or else, that species will go extinct.  To clarify, in the case of the glucocorticoid receptor, in the time it would take for those 12 mutations to be reversed, it is highly likely that OTHER mutations will arise, that could be selected for, or, at least, not selected against.  These other mutations could either provide an alternative solution to the problem, or prevent the gene from ever getting back to the way it was.  This, however, does NOT mean that the glucocorticoid receptor can't evolve into a protein that again interacts with glucocorticoids AND aldosterone, but if this condition is selected for again, then it will likely be a different set of mutations that arise and are selected for to confer this functionality, or, if some workaround can't be found, and aldosterone sensitivity is correlated with increased fitness (more reproduction) then those organisms will eventually die out as they are selected against.
To extend the rearranging your furniture analogy, say you wanted the bed to be along the wall where the door to the room is located.  In order to put the bed where you wanted, you would have to block the door with the bed, thus you remove the function of the door as an entry/exit to the room.  Now, if you really want to keep things this way, you have to find another way in and out of the room.  Perhaps you decide to use the window as a door from now on (assuming your room is on the first floor), or maybe you knock out a segment of another wall and put in another door.  You didn't reverse the path of events that got you here (i.e. move the bed back to where it was and use the old door), you found a new way to achieve the same function: being able to enter and leave the room.  Nature is replete with examples of how organisms have evolved new and different ways to accomplish the same, or similar functions, this is called convergent evolution, and it usually happens when two or more different species evolve similar parts to accomplish a function that they both need to survive and reproduce (e.g. bats and birds both evolved wings and flight, though by different means).  It is not unreasonable to hypothesize that, should selective pressures change, a single species could revert to a similar body plan or set of proteins to accomplish an "ancestral" function in a similar fashion.  Thus, it may still be possible for evolution to "go backwards", but if it ever does, it won't to take the same path to get there, at least not at the level of the genes and the proteins.