> The problem was that some FFI initialization is not brought along with
> change sets properly.

I have not experimented with FFI yet.  My understanding is that while it provides a nice in-Squeak interface to external libraries, each call is somewhat (is it significantly) slower than comparable interfaces implemented in pluggable primitives.  Does your package involve calls to FFI functions in an inside loop?  If so, perhaps a pluggable primitive that in turn (dynamically or statically) links to the external library might be better?

On the patent side, I note that this is still risky, that is to the extent the patent is valid.  Is there any word how the FreeType/Apple negotiations are proceeding?