well - there is one more explanation. In reading about seismometer vaults they talk about problems you can have with differential expansion of materials. This causes stress to build up which is eventually relieved by a slip between the materials which makes quite a noise in the seismometer. The frequency of the shock wave could be much higher than the normal range of the seismometer and thus not reproduced clearly. This problem is one reason to use a special concrete mixture - no large aggregate and no rebar. My concrete vault uses this technique: only sand mix with reinforcing fibers rather than rebars. However, the pad that it sits on uses conventional concrete and is attached to a large underground boulder with some rebars cemented into holes drilled in the rock. The rock and the concrete no doubt have different coefficients of expansion. So this could be the problem - a hint might be that it is associated with large prolonged swings in temperature.
As a test today I walked out to the vault and struck the side of the enclosure (near the bottom) with a small wrench. This is the result:
you can see my footsteps
Here is another view:
This does not look all that different than what we have been seeing once in a great while over time.
So a single-short-sharp-shock with frequency components above 50 Hz certainly could be one cause.