There are several forms of ĕsse and ēsse (= edere) that only differ by the quantity of the initial vowel, perhaps the most common one being ĕst/ēst. How do we know this difference in quantities?
There is a question on knowing hidden quantities and the answer gives several mechanisms. I would be happy to hear how the two quantities are known in this specific case.
Servius (4th cent. CE) in his commentary on Virgil writes:
sānē EDO habet et rēctam, sed antīquam dēclīnātiōnem, ut 'edō edis edit', et anomalam, ut 'edō ēs ēst': quārum secunda et tertia persōnae longae sunt propter differentiam 'sum es est'
(edo famously has a regular (but archaic) inflection, as in edō edis edit, as well as an irregular one, as in edō ēs ēst. In the latter, the 2nd and 3d persons are long for the sake of distinction from sum es est.)
Velius Longus (2st cent. CE) refers to a certain Nisus as advocating spelling a single S in comēsse and cōnsuēsse on the analogy of cāsus, causa < cāssus, caussa, which were no longer spelled or pronounced with a geminate:
Nimiae rūrsus ēlegantiae sectātōres nōn arbitror imitandōs, tam et sī Nīsus auctor est ut COMESE et CONSVESE per ūnum S scrībāmus et dīcit ratiōnem, quia iuxtā prōductam vōcālem <gemināta> cōnsonāns prōgredī nōn soleat
(On the other hand, I don't think we should take after those who pursue elegance too eagerly, even though Nisus advocates writing comēsse and cōnsuēsse with a single S, with the reasoning that a double consonant doesn't normally follow after a long vowel)
Donatus' (4th cent.) commentary on Terence's Andria (line 81):
et prōductē legitur ESSET, ut 'cibum capiat', et correptē, ut alibī
(ESSET is read both with a long vowel, as in '(s)he should eat', and with a short one, as in the other case)
