Octopodes is clearly incorrect in English, because if the plural were octopodes the singular would be octopous. All masculine delta-stem and tau-stem Greek nouns end up with -ous in the nominative singular, because the dental always drops out when sigma (the nominative singular masculine ending) is appended to it, and the omicron lengthens to compensate for its loss, resulting in omicron upsilon. You can see the same thing in pous (foot - the root is *pod) and many other words.
English only uses the plural form from the language of origin when the singular form is also visibly retained from the language of origin. The attested English singular, octopus, is rather obviously a complete mismatch for octopodes. Therefore, octopodes is invalid in English.
If the word "octopus" comes to English via Latin, then "octopi" is valid. Because the -us ending is clearly visible in the English word, the -i plural would therefore also be valid. The fact that the root originated in Greek is unimportant -- it's the language that we got it from directly, which is still visible in the singular form, that matters. If there's any foreign language of origin visible in the singular form here, it's Latin.
Note too that "octopi" would still be valid even if it were never attested in Latin, if the singular form ends in -us in English because of the Latin pattern. If the English singular form uses the Latin ending *because it's the singular ending in Latin*, then the Latin plural is valid in English as well. It's the presence or absence of the singular element, its visibility in the English singular form, and (rather importantly) the fact that *that's why the singular form is spelled that way in English* that governs the validity of directly importing the plural form as well -- to match, as it were.
However, there are plenty of words ending in "us" wherein the "us" is not in fact a Latin ending, e.g., "bus". If the singular form in English does not retain any visible singularizing element from another language, then the standard regular English declension (add -s or -es) is used (unless the word fits into the null plural declension like "deer" or "sheep"). It is arguable that "octopus" falls into this category, because no Latin origin has been documented for it. \
Bottom line: if the -us in "octopus" is a corruption of the Greek ending -ous, then the only correct plural is "octopuses". If the -us in "octopus" is patterned after the Latin singular ending, then the Latin plural ending is valid in English. The form "octopuses" might still be valid as well, of course, but educated people tend to prefer foreign-form plurals when they are available. If we could only prove whether the -us in the singular form "octopus" is or is not due to Latin influence...