The Proto-Indo-European "third laryngeal", *h3, is often assumed to have been a voiced sound based on the fact that some reflexes of the "drink" root *peh3- appear to show voicing assimilation of p to b: e.g. Sanskrit pibati < *pi-ph3-, Latin bibit similarly with further regressive assimilation.
But whenever I've seen this argued, this "drink" verb is the only piece of evidence presented. Are there other words that show assimilatory voicing of a consonant next to *h3? It's true that *h3 seems to have been the least frequent of the laryngeals, but still, you'd think that more than one data point should be adducible for this process.