Looking up the etymology of punch in the OED3 out of curiosity, to see if anything interest had been discovered by the OED editors during their run through the letter P, I came across this:
Apparently < Sanskrit pañca (also with vernacular pronunciation pañc) in pañcāmṛta, lit. ‘five nectars (of the gods)’, combination of five medicaments, so called on account of its five ingredients, i.e. milk, curd, butter (probably ghee), honey, sugar (probably molasses) < pañca five adj. + āmṛta, plural of āmṛt nectar ( < the same Indo-European base as ambrosia n.); perhaps compare also Sanskrit pañcapātra set of five glass bowls for libations; the Sanskrit collective compounds would usually have been written and thought of as separate words.
āmṛta is parsed as the plural of what looks like a compound *āmṛt with, as a final member, a noun stem *mṛt- with the -t- extension beloved by Sanskrit for making vowel-final roots inflectable as nouns. Apparently several errors have been made here, perhaps through confusion caused by the pronunciation of Sanskrit tatsamas in Hindi, by inattention to the proper resolution of sandhi, and by carelessness in the use of stem-forms for the citation of substantives. First of all, the word for "nectar conferring immortality" is amṛtam, with a short initial a-, not a long ā-. From the synchronic perspective, this a- is the a privative of Sanskrit, although the compound amṛtam is doubtless the reflex of a formation of Proto-Indo-European date, and the -mṛtam is likewise transparently from the root mṛ-, "to die."
Also, the OED3 citation of etymon as simply pañcāmṛta is confusing for the Sanskritist. They ought to have written pañcāmṛta- or used a fully inflected form as the citation form—the nominative for a masculine or feminine substantive, and the nominative-accusative for neuter substantives, as in the present instance. The compound pañcāmṛta- is used in the later literature both as a neuter singular pañcāmṛtam and a neuter plural pañcāmṛtāni according to Monier-Williams. A collocation of pañca and amṛtam "nectar" is not attested in the Rg Veda, where the word does amṛtam indeed make the archaic regular plural amṛtā with long ā (not, however, *amṛta). As the OED etymology is currently written, it looks to the Sanskritist as if a putative consonant stem *amṛt makes a plural *amṛta. That is plainly ungrammatical. What kind of neuter plural ending is -a in Sanskrit?
I am ashamed to say that I was not sure I could produce a properly formed plural for a neuter consonant stem having the shape *amṛt off the top of my head. *Amṛnti? I had to slink away to my bookshelf to consult Whitney's Sanskrit grammar to see what the proper form was—indeed *amṛnti. In any case, no such stem *amṛt- (which ought to be an adjective meaning "not dying") exists, as far as I can gather from Monier Williams. What the OED meant to say, I think, is that pañcāmṛtam is a compound of pañca "five" and amṛtam, nectar of the gods, and that in rendition of this compound in the modern Indic languages the -am of the neuter is ignored.
All this aside, in the continuation of the OED3's etymology of punch, the discussion of the vowel in the word and its quality in 17th and 18th centuries is fascinating.