Are vestiges or influence of the instrumental case in any way identifiable in Latin and Greek?

Question

I believe the instrumental case was absorbed by the ablative in Latin and by the dative in Greek. Is there any way at all in which influence of the old instrumental can be seen in Latin or Greek?—outside the bare fact that the ablative and dative, respectively, can be used to indicate an instrument.

Any remains in morphology, in semantic oddities, in syntactic sequences, etc., or whatever you can think of, would be of interest. I know very little of the instrumental case, so I don't really know what to look for. Of course "no" is also a valid answer. (For comparison, I believe the typical vowel i of the Proto-Indo-European optative mood is in some way reflected in the Latin subjunctive.) [Additional example: the way dual endings survive in words like frenī, suggested by Draconis.]

Answer 1

The instrumental -φι suffix in Homeric Greek seems to be derived from the PIE plural instrumental case, which apparently still existed in Mycenaean Greek.

From Smyth's grammar (280):

-φι（ν) is often added to noun stems in Hom. to express the relations of the lost instrumental, locative, and ablative, both singular and (more commonly) plural; rarely to express the relations of the genitive and dative cases.

An example from Homer (Odyssey 6.6):

οἵ σφεας σινέσκοντο, βίηφι δὲ φέρτεροι ἦσαν.
They [the Cyclopes] were stronger in force than they and plundered them.

Answer 2

According to Frisk, the omicron-stem seems to make an omega-ending with a few adverbs, while others in the a-stem will become an eta. The consonant stems can be found in compound words, and, in Frisk on ἐκεῖ, itself a locative, an alternative form is "κῆ (Sapph.), an old instr." Frozen caseforms look like other normal-looking declensions so you can only see faint glances at them in correlatives. Middle Liddell describes πῆ. In another entry I don't remember, a not-Ionic-Attic form of πῶ was an either instrumental or ablative fossilized interrogative correlative. Others on Wikipedia's table of correlatives might hint at other cases. Plurals are answered as by others above as marked with -φι(ν), though by Homer's day, they were used in singular as well, and possibly once had ῳς(?)** in second declension plural. The PIE and Protohellenic charts I can see make the instrumental in the singular look like the nomin-accusati-vocative dual, with an *h₁ suffix, making an oxytone (when not recessively accented) long thematic vowel, except with the long α freely turning into η (The Ionic raised ᾱ and Attic reversion).

The charts also lead me to guess, because of its deriving PIE from ancient Indian languages, that the Dual is PIE *abʰām, which would make an unattested Greek Instrumental Dual *-όφην, *-άφην that was never used or didn't exist.

** EDIT: per Osthoff's law, the *ῳς above would have become the familiar instru-dati-locative -οις, being a long diphthong in a closed syllable.