Fast is a poor word to talk about benchmarks between languages. Memory, throughput, latency are the things we should be talking about. There’s also trade offs where predictable scaling of latency under massive load is important (Erlang, Elxir), or spending less memory and starting up quickly in order to scale horizontally is important (Golang), or speed of development is important (PHP, Python), or correctness is important (Haskell, Rust). Java is rarely the winner in any area in favor of being well rounded, but as we should be using tools that meet at least the broad requirements it is always possible to make a better choice than Java.
> I’ve seen apps run for years straight.
This is why Java is dated. I am an Erlang developer, which is known for being hot reloadable and running forever. It turns out this model is actually juxitposed to Kubernetes and severless, which both presuppose letting processed crash at the host level and starting and stopping quickly. This also gets into the tooling argument because profiling in Java would rather the VM be long lived.
> The tooling.
I too have an operating system with great tooling, it’s called Linux. My knowledge of tooling is more broadly applicable, I can introspect any binary and attach probes to the kernel with bpf. Profiling for a VM is table stakes because you’re on your own. Not only do you have to figure out what’s going on in the JVM you also need to duplicate your effort to figure out how the JVM is running on the host.
> Java OO is fine.
OO in general is not fine.