An InputStream ties up a tiny kernel resource, a low level file handle. In addition, the file will be locked to some extent (from delete, renaming), as long as you have it open for read. Lets imagine you didn't care about the locked file. Eventually, if you need to read another file, and open it with a new InputStream, the kernel sequentially allocates a new descriptor (file stream) for you. This will eventually add up. If it is a long running program then it is just a matter of time until your program fails.
The file descriptor table for a processor is typically of limited size. Eventually the file handle table will run out of free slots for the process. Even in the thousands, you can still easily exhaust this for a long running application, at which point, your program can no longer open a new file or socket.
The process file descriptor table is as simplistic as something like:
IOHANDLE fds[2048]; // varies based on runtime, IO library, etc.
You start with 3 slots occupied (STDIN, STDOUT, STDERR). Also any network sockets and other types of IPC will use a slot in the same table. Fill that up and you have performed a denial of service on your program.
All that is nice to know; how best to apply it?
If you rely on local objects to go out of scope, then its up to the Garbage Collector, which can reap it in its own sweet time (nondeterministic). Don't rely on the GC, close the streams explicitly.
With Java, you want to use try-with-resources on types that implement java.lang.AutoCloseable, "which includes all objects which implement java.io.Closeable" per the docs: https://docs.oracle.com/javase/tutorial/essential/exceptions/tryResourceClose.html
With C#, the equivalent is a "using" block on objects that implement IDisposable
java.io.ByteArrayInputStream
does not use any file descriptors, and hence closing it is a no-op, see docs.oracle.com/javase/8/docs/api/java/io/… – Saarinen