The question Are there exceptions to the Latin stress rules? has an answer by Joel Derfner saying that the first-person singular perfect forms dormiī, audiī, veniī (for dormīvī, audīvī, venīvī) have penult stress, and an answer by Alex B. saying that contracted perfect forms like audīt from audīvit or disturbāt from disturbāvit are supposed to have had ultimate stress.
That sounds plausible to me, but I'm confused because at one point in "Aspects of the Phonology and Morphology of Classical Latin" (2016) Andras Cser says that neither i in abiit was stressed (126.96.36.199; p. 121). Cser analyzes eo and its compounds as having a perfect stem that underlyingly ends in ī- (p. 117), which might be relevant, but I'm not sure whether it is commonly accepted that audii(t) and abiit had different stress patterns.
Also, Leumann 1977 says "diese Akzentuation perf. audít cupít fumát gegenüber praes. aúdit usw. ist sicher von Grammatikern differentiae causa für Dichterinterpretation in der Schule festgelegt (vgl. § 237, 1d); also faßten sie audít als verkürztes audīvit, d. h. als kontrahiertes audī́it, nicht aúdiit", which if I understand correctly implies that aúdiit, with antepenultimate stress, was a form that was actually used (Lateinische Laut- und Formenlehre, p. 601).
This made me wonder how we would even know where the stress fell in such words. The "remote past tenses" in Romance languages seem to indicate stress on the i (e.g. Italian "udii" seems to have penult stress), but the position of stress in Italian and other Romance languages isn't always the same as the position of stress in Classical Latin. Is there any other evidence for the stress patterns of these perfect forms in Classical or Pre-Classical Latin?