The situation with VOT of word-final plosives is that you can measure it, although in some cases the value might be infinite. VOT isn't a thing that exists or doesn't exist, it's a measurement. The main reason why word-final VOT isn't a big topic of conversation is the lack of voicing after the stop (hence the infinite VOT), hence /d/ and /t/ are identical w.r.t. VOT (even when not completely identical). Therefore, measuring word-final VOT is mostly a useless exercise. However, if you rephrase the question and ask about phrase medial but word final, then there is something measurable. The hypothesis would be that word boundary has an effect on VOT in comparable segmental contexts, i.e. VCV versus VC#V. This is a slightly sensible project, but presupposes an understanding of "simpler contexts" so that relevant variables can be controlled. There are studies that look at the effect of different phrasal positions on word-initial VOT which indicate that a "stronger" phrase boundary leads to increased duration of consonants, from which follows increased VOT (comparing CV, V#CV and V##CV).
However, remember that VOT is a phonetic exponent of phonological voicing, and when the primary question is "what are the cues to the voiced / voiceless distinction", because other cues are more important for final stops, the influence of word boundary on phrasal VOT is less important. It may be that the question has been directly studied for English. It has been studied for Eastern Armenian, in Hacopian 2003 "A three-way VOT contrast in final position: data from Armenian" (Journal of the International Phonetic Association). That article provides various measurements and analysis of interactions between variables, including VOT. I suspect that the dearth of studies on Armenian phonetics and thus our knowledge of basic facts of Armenian is a motivating factor, and the over-abundance of phonetic studies on English is a deterrent.
