You need to map the circuit more carefully. Slapping in a tester doesn't really tell you if it's wired correctly. Noting what goes out when it trips is an incomplete mapping effort.
A GFCI+receptacle combo can be tripped by any of the following:
- A device plugged into one of its sockets, which has a ground fault. Presumably either this device has a 3-prong cord, or energy is leaking to something grounded (e.g. the apocryphal dropping the hair dryer in a sink of water).
- Something wired into the LOAD terminals has a ground fault, and the GFCI is correctly detecting it.
- The GFCI device itself is malfunctioning.
Nothing is plugged in, so we can check that off the list. You have no idea what is attached to the LOAD terminals, so that's a big fuzzy question mark. Then of course the GFCI device itself could be failing.
Now you say you think the LOAD terminals feed another GFCI device. Ohhhhh...kay. You can do that, but you're basically playing a "Yo dawg" joke on yourself. The protection is entirely redundant. The downline GFCI adds no additional protection. Further, if there is a ground fault downline of both GFCIs, both of them will trip. Some GFCIs can be "peculiar" about the sequence they are reset in. This can drive you nuts. I would eliminate that arrangement if it's feasible to do so.
All receptacles protected by an upstream GFCI should have a "GFCI Protected" sticker or label.
Some people have a big problem with the idea of simply removing a load from the LOAD terminals without knowing exactly what it is. I somewhat agree, but I would say that if you didn't know it was protected (and there was no label), where's the loss? If you really want to, you can get a GFCI tester and do an exhaustive search of every outlet that loses power when that breaker is turned off, and see whether it loses GFCI protection if you remove it from the LOAD terminals.