Two sources that I've come across indicate a long vowel /aː/ in the first syllable of the word vallum 'palisade wall' (that is, vāllum).
This form is given in The Latin Language, by Charles E. Bennett, 1907 (p. 31), and also in Lingua Latina Per Se Illustrata, by Hans Ørberg, (p. 89). Although not especially recent, these both seem generally reliable resources to me, so I would imagine the authors had some fairly good reason(s) for supposing the existence of a long vowel in this word. Since the length of the vowel cannot be deduced from the word's scansion or from the reflexes in any language descended from Latin, apex use seems like the only possible indication of vowel length in this context. On page 66, Bennett provides the following list of inscriptions that mark this vowel with an apex:
VÁLLÁRI, CIL. ii. 4509; also VÁLLIVS, VÁLLIA, CIL. xix. 4039.
(I haven't done any further work to check this evidence.)
On the other hand, the more recent work The Reflexes of the Proto-Indo-European Laryngeals in Latin, by Peter Schrijver, 1991, gives vallum (which Schrijver derives from the Proto-European form *u̯h₂lso-) as having a short /a/ (pp. 170-171). Schrijver presumably has some reason aside from the etymology of the form for supposing the word to have a short vowel, since the form is cited in support of Schrijver's argument that PIE #RHC- regularly became Latin RăC-. Can anyone tell me what evidence Schrijver could have been thinking of?