I’ve rewritten one of the paragraphs:
We can use a simple word analogy here. Start with the word UNDERKING and add a typo to make it UNDERLING. An intelligent mind will immediately see the problem and correct the mistake. But what about unguided chemistry? First, the system would have to recognize there is a problem, and biology has no way to do that. If an organism dies because of a genetic spelling error, it is dead and cannot correct the problem. If an organism lives with a spelling error, it does not know that a problem exists. If the organism got really lucky and a back mutation happened that fixed the wording, all would be well. But what are the chances? It does not know there is a problem in general, and it cannot know which letter is the problem even if it knew there was a spelling error in that one word. Would another mutation ever be expected to fix it when there are so many other possibilities, like UNDERWING? And what would happen if a few more mutations happened before the fortuitous back mutation arose? We might go from UNCOCKING to UNCORKING, which is unhelpful. Back-mutations are only helpful if the original context is preserved.
English words are far more interconnected than creationists using them as examples realise. As noted here, almost half of nine-letter words have single-letter changes to other words. Note also that in words every letter is fixed, which isn’t the case in genetics where variation is possible with little or no effect.
There’s another problem with this paragraph, though. As the last sentence makes clear, Sanford et al are still working with the assumption that any deviation from the original genome is going to be worse, even if it moves the genome closer to another peak in the fitness landscape. Why should UNCOCKING->UNCORKING be unhelpful?
And in the original example, why should the reverse mutation BOUTNHELDE->HOUTNHELDE be unhelpful? First, it reduces the number of subsequent mutations needed to return to the adaptive peak. Second, Sanford’s genetic entropy is supposed to be about mutations with very small unselectable effects, so why is he assuming in this paragraph that any mutation is fatal? If HOUTNHELDE->BOUTNHELDE is one of a series of very slightly detrimental mutations, then BOUTNHELDE->HOUTNHELDE is very slightly beneficial, so it is helpful.
That section has other problems that can be highlighted simply by changing the examples used:
Worse, not all mutations have the same probability. If a T→C mutation (one of the rarer mutations we see) happens, the reverse [C→T] is more likely to occur. Thus, while we are waiting for our common C→T back mutation to happen, other T→C mutations won’t be occurring in the near vicinity.
That UNDERMINES the passage and UNDERLINES the problem.
Finally, Sanford’s comment at the end of the section serves only to illustrate that Sanford hasn’t understood the claim, and has nothing but assertions.