Projects that consider "testability" as a requirement tend to produce systems that are perfectly testable, but this is orthogonal to whether the architecture makes sense and can be composed with something else.
Actually sometimes being easy to run tests on harms the code, because the code grows its surface just to plug in more tests, or the author stops at "I can test it, so it's good enough" instead of ensuring the architecture and/or interface is so simple that it can't ever possibly fail.
i assume by "orthogonal" you mean independent, i.e. just because a system is testable it does not mean it necessarily has other desirable properties.