Megaton derives from the Greek megas and the Old English tun.
Apparatchik (Latin-Russian)
Neoconservative (Greek-Latin)
Superhero (Latin-Greek)
As a priggish pedant myself, I resent that insinuation: I certainly haven’t abandoned the ancient prohibition against mixing languages to form hybrid words! 
Although I don’t have a problem with hybrid compounds that are entrenched in the language already, it does grate on me when it occurs in newly-formed words.
In fact, I may already be on record around here somewhere as complaining about the term “monokini” designating the bottom half of a bikini bathing suit sold separately from the top. Not only is it a hybrid of the Greek prefix “mono-” and the Austronesian name “bikini”, but it’s a play on the Latin prefix “bi-” replaced by “mono-” to denote “one” instead of “two”.
Obviously, the word really ought to be “unikini”!
Zombie from 2008; but I’m finding it hard to resist responding, that in Hungary the “p-word” above, is often abbreviated in a maps / timetables context, to “pu.” – comparable to our similar abbreviative practice, re railway stations, of using. “Sta.”. One of Budapest’s main rail terminals is the Nyugati station; in abbreviated form, “Nyugati pu.” – which, looking rather rude, always appeals to my inner eight-year-old.
Many Latin words ending in “-us” don’t go to “-i” in the plural. That’s specifically a second decliension thing (Latin has five noun classes, or ‘declensions’).
“Trusteeship” apparently is made up from Old English “trust”, the French ending “-ee”, and Scandinavian “-ship”.
True, but first and second declension are by a good margin the most common, and fourth and fifth declension are quite rare. The vast majority of Latin nouns ending in -us are, in fact, second declension (though fourth also ends in -us, and third can end in pretty much anything).
Isn’t the third declension the largest?
I think that the third declension is the largest, but that the first and second declensions are more common. This is partly because third declension contains all genders while first and second split the genders between them, and partly because the words in the first and second declensions tend to be more common than the words in the third declension.
That said, my study of Latin is many years past, so if any classics scholars can correct me on that, I count myself preemptively corrected.
Also:
Hyperactive (Greek-Latin)
Liposuction (Greek-Latin)
Metadata (Greek-Latin)
Polyamory (Greek-Latin)
Geostationary (Greek-Latin)
Pseudoscience (Greek-Latin)
Telephoto (Latin-Greek)
Some non Greek/Latin examples:
Nightgown (Old English-French)
Uberleet (German-French)
Refusenik (Latin/French-Russian)
Beefsteak (French-Norse)
Miniskirt (Italian-Norse)
Prealgebra (as a school subject) (Latin-Arabic)
Awkward (Norse-Old English)
Egghead (Norse-Old English)
Flatworm (Norse-Old English)
Keelhaul (Norse-Old English)
Skiptrace (Norse-Latin)
Zeroize (Arabic-Greek)
Sodapop (Arabic-Old English)
Biohazard (Greek-Arabic)
Crawlspace (Norse-Latin)
Blockhead (Dutch-Old English)
Bathrobe (Old English-French)
Battlecruiser (Latin-Dutch)
Sketchbook (Dutch-Old English)
Sodomize (Hebrew-Greek)
Chemotherapy (Latin-Greek)
Chlorofluorocarbon (Greek-Latin-Latin)
Spectrometer (Latin-Greek)
Cyberspace (Greek-Latin)
Oxygen appears to be Greek and French. Helium is obviously Greek and Latin. Hell, the periodic table is loaded with compounds (einsteinium??).
and then there are evil mutations like “barista”
Perhaps, but very few third declension nouns end in -us (in the nominative). The point was that the vast majority of -us nouns are second declension, and so end in -i in the nominative plural.
Both the “bar-” part and the “-ista” part are from Latin, no? True, they travel via different routes before getting joined up in Italian and then loaned into English, but they are basically Latin.
On the contrary, there are no compounds in the periodic table at all. 