Some implementing classes in the FFM API use Streams, lambdas, and other constructs that cannot be used early in the init process. Some of these constructs also entail spinning bytecode which will have a detrimental impact on startup time. It is likely that, over time, the FFM API will be used to replace JNI and other native operations in the JDK internals.

Therefore, we should explore replacing said constructs with more primordial variants. This would provide two main advantages:

* The ability to use FFM earlier in the init phase
* Faster startup

The tradeoff would be giving up some readability.