ADDITIONAL SYSTEM INFORMATION :
Latest OpenJDK 11 on 64bit Linux, can also reproduce with latest OpenJDK 8.

A DESCRIPTION OF THE PROBLEM :
Quoting the Javadoc for LSSerializer: "[...] Any characters that cannot be represented directly in the output character encoding are serialized as numeric character references [...]"

Serializing text nodes that have a codepoint represented in 4 bytes in UTF-8 (for example <Tag>🚩</Tag>) produces a numeric character reference. Following the javadoc, this code point however is directly representable and should therefore not getting serialized as a numeric character reference. I haven't tested this with 2 byte code points yet.

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Parse an XML document with a text node containing a 4 byte UTF-8 code point into a DOM document, serialize it back.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
Text nodes should look exactly the same.
ACTUAL -
Original text node has the code point as-is, produced text node has the respective character reference.

---------- BEGIN SOURCE ----------
import org.w3c.dom.bootstrap.DOMImplementationRegistry;
import org.w3c.dom.ls.*;
import org.w3c.dom.Document;
import java.io.*;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;

public class Main
{
  public static void main(String[] args) throws Exception
  {
    if (args.length == 0) {
      System.out.println("Please enter a file name");
      return;
    }
    
    Path filenameIn = Paths.get(args[0]);
    Path filenameOut = filenameIn.resolve(".out");
    
    try (InputStream in = Files.newInputStream(filenameIn);
         OutputStream out = Files.newOutputStream(filenameOut)) {
      
      DOMImplementationLS impl = (DOMImplementationLS) DOMImplementationRegistry.newInstance().getDOMImplementation("LS");
      LSInput input = impl.createLSInput();
      LSOutput output = impl.createLSOutput();
      LSParser parser = impl.createLSParser(DOMImplementationLS.MODE_SYNCHRONOUS, null);
      LSSerializer serializer = impl.createLSSerializer();
      
      input.setByteStream(in);
      output.setByteStream(out);
      output.setEncoding("UTF-8");
      
      Document doc = parser.parse(input);
      
      serializer.write(doc, output);
    }
  }
}
---------- END SOURCE ----------

FREQUENCY : always