I think the LGPL is becoming particularly irrelevant. I’d like to use GEOS (LGPL 2.1) in a macOS app but the risks are high as you point out. I think it’s going to be significantly less risky long run to rebuild GEOS in Swift. The complexity and lack of clarity in the licensing is a real burden.

I don’t think this is the spirit of the LGPL - it’s just code signing prevents the use of dynamically linked libraries in the way LGPL intended.