In Homer, the form ᾔδη "he knew" (3sg. pluperfect of οἶδα) scans as if it began with digamma. This is most evident in Iliad 1.70, where the first syllable scans heavy:
ὃς ᾔδη τά τ᾽ ἐόντα τά τ᾽ ἐσσόμενα πρό τ᾽ ἐόντα
Likewise Iliad 6.351, which also begins ὃς ᾔδη.
Elsewhere (Iliad 2.38, 2.213, 5.64, 5.326, 11.741, 13.355, 13.674, 19.155, 20.466) ᾔδη is preceded by a short vowel which is not elided, which also accords with an original initial digamma. (I haven't looked at the occurrences of this form in the Odyssey, or at forms other than the 3sg.)
Of course it's well known that the verb οἶδα was originally ϝοῖδα. But I can't see how to get a digamma-initial form in the pluperfect.
The pluperfect stem of οἶδα is usually ᾐδ-, as here (though there is one occurrence of a form ἠείδει in the Odyssey). The eta would most obviously seem to result from contraction between the augment and the root diphthong ει, after loss of digamma: ἠ-ϝειδ- > ᾐδ-. (Apparently, as the Odyssey form suggests, the augment was or could be long in this verb, for some unexplained reason; that's strange, but I don't think it matters for this question.) However, if there's an augment in this form, then the digamma is no longer initial. In any case, positing digamma loss with subsequent contraction in a common Homeric form seems problematic, since digammas are usually still traceable in the Homeric dialect.
Alternatively one could imagine a pluperfect formed without the augment; then the stem might be ϝειδ-, but that would yield εἰδ- rather than ᾐδ-. What we need for the scansion is a form like ϝῃδ-, but I can't see how to get that.
There's a 1999 article by Schrijver (JSTOR access required) which discusses the pluperfect of οἶδα with comparanda from Celtic and other branches and concludes that it does come from ϝειδ- and/or ἠ-ϝειδ-, but unless I've missed something (which is quite possible), it doesn't seem to address the fact that neither of these forms would regularly give the metrically required ϝῃδ-.
So, assuming ᾔδη did have initial digamma, how do we get the long diphthong? If it didn't have initial digamma, why does it consistently scan as if it did?