> As locatives typically count as adverbs, and may become fully
> lexicalised as such (cf. *h2ant-i 'in front' or *per-ut(-i) 'last
> year'), I wonder if the *dHi element isn't itself a fossilised locative
> (of the root noun corresponding to *dHeh1- 'put, place').
If so, IE *-dHi would have to be *-dHh + *-i > *-dHi, since it's easy
to get *e out of a vocalized *h but an *-i is a little harder. It
seems more optimal to reconstruct an IndoTyrrhenian postposition *dei,
signifying specifically "in". Since both Tyrrhenian and IE languages
seem to have the same endings, any origin in verbs would have to be
fairly early, if one accepts IndoTyrrhenian.
However, conceivably within my idea, we could have adverbs, as you say,
derived from verbs such as, let's say, *ai (< ITyr *?ei-) and *dah (<
ITyr *deh-) with neutral unstressed schwa (written as *a here). In
Tyrrhenian *-h disappears in order to yield *-ta: and then *-te while
becoming *-dHe in IE (found in mediopassives, btw). If we add locative
*-ai to MIE *-dah before Syncope, we'd have extended *-dah-ai, becoming
eLIE *-dHi (since unstressed *-dah-ai > *-dHhi > *-dHi).
> Is it merely a phonetic process, or, as in the case of "intrusive [r]"
> in non-rhotic English, a lingering trace of an otherwise lost segment?
> In other words, does the locative *-i derive from older *iN?
But intrusive-r is seen in places where there is no historical [R], as
in a British pronunciation of "China is in Asia" ['tSajn@... Iz In 'EjZ@].
That in itself can't be indicative of anything.