Others will have to investigate the details and establish the mapping between the incoming data and the behavior.
In the meantime:
Which alpha mode do the glTF actually use? Is it OPAQUE or BLEND?
The point here is: Even though the message
“Alpha cutoff is supported only for ‘MASK’ alpha mode.”
is only a WARNING, it sounds like something that could explain the behavior: If the alpha mode is BLEND, then this could - depending on some lower-level details and possible default values - cause the tileset to be rendered ~“with a (completely) transparent material”, which might explain why nothing is visible.
(Note: Although the BLEND mode should work in principle, the fact that the warning is emitted might hint at some form of ~“inconsistency” in the data that might cause the unexpected behavior here)
Going out on a limb, you could try
- assigning a custom material to the tileset (probably roughly as described in this post) and see whether the tileset is visible then
- if the glTFs actually use
BLEND, try out whether it works when they useMASK(orOPAQUE) instead