Linux is highly readable because Linus imposes readability on it through sheer force of will, and because they are willing to sacrifice driver API on the altar of clean code. The result is a very efficient and readable kernel that hardly anyone uses in the environment with the most complex and heterogenous hardware i.e. desktops and mobiles. Android doesn't really use Linux these days. Since some years it has developed its own stable driver ABI that bypasses the kernel for most things.
If all Java codebases were run by a dictator-for-life who could sacrifice trifling details like adoption in order to get the most readable code possible, you'd find you'd like them a lot more too. And a few such codebases are actually like that! Look at the source code of the JDK itself sometime (not the bits by Doug Lea though), or Guava, or really quite a few open source Java projects have readable and nice code.
Java became overcomplicated by frameworks and design patterns, so has always felt bloated and slow. But has worked best for business logic. Though is maybe behind COBOL thus far, and tends to hide/obscure logic. No two java designs looks the same, scales well or converge to clear consensus. The language did lower the bar for what cheap programmers could accomplish faster though.
What vendor lock-in? The whole thing is open source from the compiler to the runtime - https://github.com/dotnet
If anything Java is more locked in as the most popular and 'official' runtime/devkit (Oracle Java/JDK) is completely closed source.
Actually, Java has great tooling, JVM is very nice, it has great potential for high performance code generation. It has everything for debugging. But at the end of the day, language is driving people to write bloated software.
I just wonder, how come e.g Linux kernel code 100 times more readable than any project in Java? More importantly, how did we come to this point that we accept bloated/unreadable code in the name of higher/better languages?