I don't think "JVM is more popular" argument does justice to Java (and Kotlin) strengths. With this reasoning, you could also say "C++ is more popular for systems programming" but it doesn't stop developers from switching to Rust, Zig or even C# as a wider scope and easier to use language that has gotten good at it.
Nonetheless, you could make this argument for select Apache products, but that's Apache for you. It does not hold true for the larger ecosystem and, at the end of the day, quantity is not quality, otherwise we would've all been swept by Node.js :)
Same applies to "packages that bundle native libraries".
First, they are always maintenance-heavy to manage with ever growing matrix of platforms and architectures. Just x86 alone is problem enough as all kinds of codecs perform wildly different depending if AVX2 or 512 is available vs SSE4.2 or even SSE2 without EVEX. Now add ARM64 with and without SVE2 to the mix. Multiply this by 2 or 3 (if you care about macOS or FreeBSD). Multiply linux targets again by musl and glibc. You get the idea. This a worst-case scenario but it's something Java is not going to help you with and will only make your life more difficult due to the reason below.
There is also an exercise in writing JNI bindings. Or maybe using Java FFM now which still requires you to go through separate tooling, build stage, deal with off-heap memory management API and still does not change the performance profile significantly. There's a reason it is recommended to avoid native dependencies in Java and port them instead (even with performance sacrifices).* Green Threads will only exacerbate this problem.
Meanwhile
using System.Runtime.InteropServices;
[DllImport("libc", EntryPoint = "putchar")]
static extern int PutChar(int c);
var text = "Hello, World!\n";
foreach (var c in text) PutChar(c);
since C# 2 or maybe 1? No setup required. You can echo this snippet into Program.cs and it will work as is.
(I'm not sure if binding process on the ole Mono was any different? In any case, the above is a thing on Linux since 8 years ago at least)
Like with other false claims, the "better native packaging" one is easily disprovable. Just look at the list of supported platforms here: https://github.com/sandrohanea/whisper.net?tab=readme-ov-fil...
It even supports CoreML on iOS and macOS.
* Now applies to C# too but for completely different reason - you can usually replace data crunching C++ code with portable pure C# implementation that retains 95% of original performance while reducing LOC count and complexity. Huge maintenance burden reduction and "it just works" without having to ship extra binaries or require users to pull extra dependencies.