Spell it separated (as two orthographic words) or solid (as a single orthographic word)? (See discussion of solidifications here.) Two examples — of different types — that have come to my attention recently: for ever (vs. forever) and all most (vs. almost).
1. for ever. From “The Gulf: Still rich but no longer so calm” in the Economist of 12/13/11 (p. 61):
The region’s rulers have long relied on their oil wealth as a way to keep their citizens quiet but as ripples of unrest spread, this may not work for ever.
What I said in my “Solidification” posting was:
There is currently variation in the spelling of for ever/forever ‘always’, though the solidified spelling seems to be gaining ground fast.
I didn’t say there that my own spelling of this expression is always solidified — so I noticed the separated spelling in the Economist, taking it to be a British spelling. Bryan Garner (Garner’s Modern American Usage 3rd ed., p. 370) is sterner:
forever. One word. The OED suggests a distinction – the solid forever meaning “incessantly” <she’s forever twirling her hair> and the two-word *for ever meaning “for all future time” <we’ll be together for ever>. But the solidified version has become standard in both AmE and BrE, and the two-word version is best described as archaic.
The separated spelling is the original, but solid spellings are attested in the OED from 1670 on. OED2 distinguishes the two spellings in part by sense (see Garner), but more crucially by geography, with the solid spelling characterized as American. The separated spelling is listed under the adverb ever:
in phrases, for ever (sometimes, esp. in U.S., written forever adv., q.v.): for all future time, for eternity, in perpetuity; hence (chiefly in colloq. use), incessantly, interminably. In proper sense often in strengthened forms, †for all ever, for ever and (for) ever, for ever and ay (arch.), for ever and a day (? a corruption of prec.; also hyperbolically, (for) a very long time).
OED2′s cites run up to 1878 (presumably carried over from OED1), then have a gap until 1934, with an American for ever (John O’Hara), followed by another American cite (1966, Joan Aiken) and a British cite (for ever and a day, 1967, in The Listener). The OED’s sampling of usage mostly goes back over a century ago, as does its characterization of the contexts favoring the variants.
COCA has a huge disparity in favor of the solid spelling in the U.S.: 18,467 hits for it, vs. only 361 for the separated for ever (and a great many of those are historical quotes, or come from archaizing genres, or are typos).
OED2 on the solid spelling:
The phrase for ever (see ever adv. 5b), written as one word. Chiefly U.S. exc. in sense ‘incessantly’. [cites from 1670 through 1875; again, presumably carried over from OED1]
In the draft additions of Dec. 2002:
hyperbolically. A very long time, esp. a (seemingly) inordinate amount of time. [first cite in 1861, from the Atlantic Monthly, with "There we sit waiting and waiting forever".]
In any case, orthographic change seems to have been rapidly running its course from separated to solid, though the separated variant still occurs, primarily in BrE.
2. all most. A very different story here. The etymology does begin with a combination of separate words:
[OED2] < all adj., n., and adv. + most adj. = mostly
The combination was already mostly solidified in OE (as ælmæst, with variants) in the obsolete sense ‘mostly all, nearly all; for the most part’, and was apparently always solidified in the sense ‘very nearly, wellnigh, all but’. So all most in modern writing is something of a surprise. But it’s there, with considerable frequency (though not, so far as I can, in polished writing).
Unable to boot into recovery – tried all most everything (I think) (link)
[Eyehategod song] Lack Of All Most Everything (link)
Users Interested In all most everything [on ChristianMingles.com; writers of the entries use almost] (link)
Steiner Sports 40% All Most Everything Sale (link)
Even excluding dupes and a fair number of cites from non-native speakers, there’s a huge number of cites, for all most everything, and also all most everyone, all most anyone, all most nobody, all most all, etc.
Now contrast for ever/forever with all most/almost. Both are fixed expressions, people probably use them without reflection on the connection between their form and meaning. But you can fairly easily make sense of for ever/forever as composed of the durational P for and a time-span N ever, while all most/almost is much harder to rationalize. For ever will be inclined to solidify, but that’s just because fixed expressions with function words in them show a general tendency to follow that path. Why should almost desolidify?
I think the key here is the fixed expression very often spelled ALRIGHT, which is as opaque as almost, and like it is spelled with AL rather than ALL. The spelling ALRIGHT has been the subject of intense disapproval, from sources that insist on the normative spelling ALL RIGHT (see MWDEU‘s entry on the pair). People have been told that ALRIGHT (despite its considerable frequency) “is correctly spelled” ALL RIGHT, and that advice could then carry over to the parallel ALMOST — parallel in the spelling AL- for the first part and in the recognizability of the second part (-RIGHT, -MOST).
So advice about the spelling ALRIGHT (whose only defect is to have solidified later than almighty, almost, alone, already, also, although, altogether, and always) results in the mis-spelling ALL MOST. And, in fact, some others as well: You are all ways on my mind, I have all ready finished the game, and possibly more. These are unlikely to be survivals from much earlier spellings, but are new creations, probably encouraged by ALRIGHT.