None of the example models are actually compatible with vsgclip example

[AnyOldName3]

All of the example models have something wrong with them when running the vsgclip example. For example, the teapot clearly reads from uninitialised memory if run without validation layers enabled, and simply refuses to draw otherwise, and the OSM map is simply way larger than the clipping sphere. The most compatible of the models is lz.vsgt, but even that has problems with the billboarded trees, which all end up stacked in the same location at (presumably) the world origin.

As the pipeline replacement to swap in the clipping-enabled shader needs the original pipeline layout to be compatible and the shader's behaviour to be roughly the same (minus the clipping) it would make a degree of sense to pick replacement state compatible with the entirety of at least one of the sample models.

[robertosfield]

I think you are expecting too much from a little example, it's not intended to be a general purpose tool, just a simple example. One of the challenges with making a custom shader work with random models is it's likely to struggle with incompatibilities.

Perhaps just creating a synthetic model might avoid these issues.

[AnyOldName3]

It'd be pretty simple to tweak it so it works with the teapot example, and then it'd work properly with one of the test models. I don't think that one situation where the example works is too much, or that zero is enough.