Hmm I am not sure as I’m not a power user of VESTA. It might have something to do with VESTA’s bonding algorithm? It might be worth manually checking the distance of the H site to the nearest sites in pymatgen/python and ensuring the “bonds” you expect have distances that would normally be shown in vesta as “bonds”.
i.e., if your atom is for some reason not equidistant from at least 2 near sites, it may not show as bonded?
Alternatively, IIRC vesta has options for displaying bonds depending on the unit cell or something, it might be worth messing with those controls as well