This Java code will hold/occupy file descriptors after it runs. After using Reader, I should close it. (I once thought that using the .lines approach would close the stream, actually it doesn`t close stream.)

private static String getContent(String filePath) throws IOException {
    BufferedReader reader = new BufferedReader(new InputStreamReader(new ClassPathResource(filePath).getInputStream(), StandardCharsets.UTF_8));
    return reader.lines().collect(Collectors.joining(System.lineSeparator()));
}

Then I run this loop, try to run above code 20,000 times:

public static void main(String[] args) throws InterruptedException  {
    for (int i = 0; i < 20000; i++) {
        getContent("test.json");  // assume there is a file called test.json
    }
    Thread.sleep(10 * 60 * 1000l); // try to hold the execution
}

Next, I try to count the number of file descriptors the program holds by using the lsof command.

// assume the process id (PID) is 12921
 lsof -p 12921 | grep test.json | wc -l

Expected result:

20000

Actual result:

3932

Question

Why are there fewer file descriptors(3,932) than necessary(20,000)?

Is there some file descriptors being reused?

Any advice is appreciated, thanks in advance.