The Economist has demonstrated several times that it would rather publish ambiguous, awkward, or even ungrammatical sentences than permit a verb-modifying adjunct to intervene between the marker to and the head verb of the infinitival clause it introduces (see here and here for two of my discussions of the topic). Last week I obtained a robustly direct reaction from an influential staff member at the magazine's offices (I've given the details on Lingua Franca today). It stated that they would not be changing their highly conservative policy — it came close to telling me to butt out. But almost immediately thereafter, I came across a sentence that (you might think) looked like counterevidence. It was in an article about computer modeling of tsunami behavior (15 June 2013, p. 82); I underline the crucial part:
To simplify the problem, the researchers looked at what happens when a computerized wave encounters a cone-shaped island on a smoothly sloping seabed in front of a straight cyber-coastline with a beach that continues to rise smoothly as it progresses inland. These approximations allow a computer to cope with the problem, yet are sufficiently similar to many real places for the conclusions drawn from them to, as it were, hold water.
However, here the interpolated phrase as it were is not a routine modifying adjunct. It is a parenthetical interruption of the type that The Cambridge Grammar of the English Language calls a supplement rather than a modifier.
The two key things about supplements are that (i) they are obligatorily fenced off with commas or dashes or parentheses, and (ii) they can be stuck in places where nothing else can be stuck.
For example, English does not permit modifiers (the ordinary unfenced kind of adjuncts) to intervene between a verb and its object, if the object is moderately short and simple, so this is ungrammatical:
(If you're tempted to say that seems all right, it's because you're either reading reluctantly with commas around it, or managing to convince yourself that a lawnmower should be postponed for some special style reason. You won't have the same reaction if we make the object noun phrase really short: *I needed a lawnmower so I rented reluctantly one. That is surely ungrammatical enough to make the point.)
The modifier reluctantly can happily sit either before the verb or at the end of the clause instead:
Today I rented a lawnmower reluctantly.
But supplements work differently from modifiers; they get special dispensation. It is fully grammatical to say or write things like this:
Today I rented, assuming my Paypal account hasn't been canceled, a lawnmower.
So The Economist hasn't really altered its policy at all. It's just a matter of the special behavior of supplements that is relevant here, plus one other thing: there is a slight jocularity in for the conclusions drawn from them to, as it were, hold water, a deliberate calling of the pun to the attention of the reader, and it works best if "as it were" comes as late as possible, immediately before the minimal phrase containing the metaphorical reference to water.
The policy hasn't really been changed, then. Two special factors combined to override the magazine's atavistic policy on this occasion. It's basically an exception that merely proves the rule: my favorite magazine still has a usage policy which on some points is stuck in the dark ages of the 19th century.
Postscript: At Peter Harvey's blog you can read a critique of my position on The Economist’s "split infinitive" ban. Harvey says:
I would have expected an arch-descriptivist like Geoff Pullum to have noticed, recorded and accepted that very many people choose to use the language in just that way rather than laying down the law and, ahem, prescribing how the language should be used by those who teach and write to satisfy the public taste.
Well, if you want plain description, I could give you the statistics on how many modifiers separate to from the following verb and how many appear to the left. But even that wouldn't be quite descriptive enough: to describe the real situation we would need to distinguish between the examples occurring in copy-edited prose or other invigilated contexts from those occurring in freer writing by native speakers.
I'm certainly aware of the sad fact that the relative frequencies of modifier positions in much written material is warped by the fact that both authors and teachers have to take account of the widespread public prejudice against "split infinitives". And I am aware of the crisis of conscience English teachers and other language professionals face; as Harvey puts it:
I simply cannot afford to have a student come back to me with a complaint that something that I have taught or tolerated has been criticised in no uncertain manner as a grammatical solecism by a native speaker; nor can I afford to have an argument with a translation client on the same matter. As a result, I play it safe just as the Economist does.
Harvey can teach avoidance of "split infinitives" if he is confident that it will protect his students from being dinged, and he is at liberty to avoid the construction in translation jobs if he thinks the clients would complain. But both courses of action seem over-diffident to me. What he shouldn't do is tell students or clients that "split infinitives" should be avoided for reasons having to do with respecting English grammar, because that's simply false. To the extent that there are reasons for such avoidance, they have to do with bowing to the preferences of grammatically ignorant nitpickers.