scratching the itch

The questions that no one can answer often involve why.
A Japanese student asked me, “Why do people pronounce sandwich with dʒ, given that the spelling is -ch?”
Not everyone does, of course. But in my BrE preference poll I found that 53% of respondents voted for -wɪdʒ and only 47% for -wɪtʃ. Given the biasing effect of the spelling, the true figure for -wɪdʒ is probably quite a lot higher. (In AmE, on the other hand, I think we always get -wɪtʃ.)
The same alternation seems to apply to all words ending in -wich. Thus Norwich can be ˈnɒrɪdʒ, Woolwich can be ˈwʊlɪdʒ, Dulwich can be ˈdʌlɪdʒ. More generally, we could say it applies to all words with possible final unstressed -ɪtʃ. (Or perhaps not. Dunno about Harry Potter’s quidditch.)
EPD has a note at sandwich to the effect that “some British speakers use -wɪtʃ in the uninflected form and -wɪdʒ in the inflected forms of this word”. I do not know if there is any evidence for this claim.
The OED does not record anything except -wɪtʃ.
So why does this voicing happen? You can hypothesize about lenition of the affricate in this weak position, but not very convincingly. (Why does which not lenite in the same way when unstressed? Why do plosives, as in gossip, rabbit, topic not lenite, nor fricatives as in sheriff, Lambeth, palace, radish?)
It is interesting, too, that the alternation does not work in the other direction. No one pronounces -ɪtʃ in Cambridge or cabbage.
— —
I am experimenting with setting the font differently for my blog. It ought to look best if you have installed Charis SIL or Doulos SIL, and still be readable if you have not. How do the IPA symbols look? Do you prefer this to what we had before?