So it seems to have happened with "passive": the metaphorical meaning ("weak") reverted to a literal one misapplied by ignorant teachers and editors.

I think the problem is somewhat muddied here by the fact the non-technical (or perhaps we should call it quasi-technical) use of "passive" is very, very inconsistent. As Andrew (not the same one) points out, some people think that a passive clause is any clause with a form of the verb "to be" in it. Others think that "the passive voice" is any form of expression that is vague about agency (or indeed that fails to place blame where the complainant feels it should be placed). Still others think that "the passive voice" means anything that is in any way weak.

To put it another way, the problem with the "non-technical" definition of "passive voice" is that it only makes any sense at *all* if you mistakenly believe it to be the real, technical definition. The popular uses all claim to identify a particular way of using English, when they don't. When people say that a particular passage "avoids agency by using the passive voice" they clearly think that there is some particular type of language being used and that they are able to identify it.

Context is also important here. "Passive voice" is very different to "gridlock" in the sense that it isn't a word people use casually, rather it's a technical term that people attempt to use in technical analysis.

Perhaps a better analogy than "gridlock" would be "quantum" or possibly "genetic". "Genetic" has a very specific meaning in biology (it means, well, literally encoded in DNA) but tends to get used in everyday speech as a synonym for "hereditary" or "predetermined". This is all well and good, but once people start trying to make actual testable claims about the physical world that rely on pretending that the pop-culture definition is the same as the scientific definition (like, say, in pretty much every newspaper article about genetics you'll ever read), you really do have to sit down and say "no, you are using that word wrong". Which is really a shorthand for saying "you appear to be assuming that your understanding of a metaphorical idea which, confusingly, has the same name as a real scientific concept, is the same as understanding the scientific concept itself and you are now making false claims about an area of science you do not understand."

It is perfectly fine for people to use "passive" to mean "vague about agency" or "weak" or "contains a form of the word to be" but if those are the definitions you're using, you can't pass your comments of as an actual analysis of the language, any more than you can pass off your observation that you and your dad both like the same kinds of books as an actual analysis of the biochemistry.

[(myl) In this kind of discussion, there's a big difference between an appeal to authority of the form "Strunk (or Fowler or Ayres or …) said you must always (or never) do X", and an appeal to authority of the form "Dickens (or Carlyle or Austen or Eliot or Yeats…) routinely employed construction or usage X". The first is the prescription of a self-appointed authority; the second is a description of the practice of well-regarded writers.

…or countless other recent works by living authors. But a common response is to take such citations as evidence of failure to maintain standards.]

Surely it matters little what was said/written in previous centuries, if we adopt a descriptive viewpoint.

[(myl) Surely the reason is obvious. It's trivial to find current examples of anticipate used in the sense forbidden by the Telegraph's style guide — just search Google Books or any newspaper index for e.g. "anticipate that", and you'll find as many as you like. The expected rebuttal, especially from self-described "conservative" sources, is that this is just a symptom of modern linguistic and cultural degeneracy, we need to keep up standards, etc. So it's useful to point out that the usage in question is not some modern error, but has been common in elite writing for a couple of centuries.]

I wonder what the Telegraph Style Guide has to say about 'unanticipated', which apparently has been in use since the 18th century. Does it insist that the word is properly applied only to an expected event that has not been acted upon in advance, e.g., a trip to Africa or a marriage for which no preparations have been made? I don't think I've ever seen the adjective used in that sense.

The TSG's implicit prescription (it doesn't explicitly prescribe or proscribe any verbal behavior here–doesn't explicitly say one should not use 'anticipate' in this way) is not worth following, but I don't think it's prescriptivist poppycock–if by 'poppycock' we mean a claim that is wildly false. Prescriptions (deontic statements) are neither true nor false. No, the TSG's observation about 'anticipate' is descriptivist poppycock, pure and simple.

(1) a. I expect to have a lot of fun this summer.
b. # I expect to have a lot of fun at your party.
(2) a. I expect the economy to recover quickly.
b. # I expect you to recover from your cancer.
(3) a. I expect a lot more rain in Ohio this year.
b. # I expect a lot more sexual assaults in Ohio this year.

Anticipate (though it doesn't take an infinitival complement) at least avoids this subtle but potentially critical pragmatic problem. These days, I reflexively prefer it (or a whole host of other constructions) to expect.