JMH VS sbt-native-packager

Also, it has an amazing SBT plugin integration.

Take a look at sbt-jhm for doing benchmarks. Java in particular is hard to measure because of optimizations that happen at run-time. jhm runs multiple iterations and gives tools to ensure that function calls and loops that may be optimized away are kept around and tested. You may also find some cases that are faster in node.js because the Javascript V8 engine is highly optimized.

Also, if you really want to know what the performance characteristics are you should use JMH (sbt plugin https://github.com/sbt/sbt-jmh). Not sure how you are evaluating the performance but things like JVM startup and warming can make a big difference. JMH will give you a better idea of real world performance when the JVM is already started and any relevant hot code has been JIT compiled.

once you outgrow scala-cli, you should know sbt has a lot of plugins ( some might say it's ecosystem is the only thing keeping it relevant....) like sbt-native-packager which again does the heavy lifting for you

So far, we have been working mainly with projects that use https://github.com/sbt/sbt-native-packager, still, projects building fat jars should also work smoothly.

You should use sbt-jib. (You shouldn't use sbt-native-packager either, it doesn't work well in this regard: [1], [2].)

As others have stated, this is mainly due to the nature of the JVM. You can try using GraalVM however, which basically complies JVM bytecode to native-code avoiding startup time issues, sbt-native-packager lets you do this quite easily.

IMO don't take any of this advice and use https://github.com/sbt/sbt-native-packager