These are modular building blocks each designed to exercise a particular feature in FIDL. For example, the test.unknown.interactions library contains protocols that exercise the unknown interaction handling functionality.
Binding tests should be written using these protocols. Sharing the protocol definitions makes it easier to test all the edge cases, such as different binding API shapes for zero-arg and single-arg methods. Binding tests may add binding-specific concerns via additional binding-specific FIDL test libraries.
Parts of the compatibility test may be factorized into separate libraries here. Individual binding unit tests may then be written on top, breaking up the compatibility test into facets (https://fxbug.dev/42176858).
Dynamic test suite project: https://fxbug.dev/42053264.
Binding specific unit tests focus on testing the API of generated bindings. The dynamic test suite focuses on testing the ABI across different bindings. The FIDL schema behind both tend to have some overlaps. We'll need to wait for both to grow before evaluating whether sharing the FIDL schema is a good idea.