ADDITIONAL SYSTEM INFORMATION :
Reproduced with on Linux Ubuntu 20.04.3 LTS. Probably OS-agnostic?

./java -version
Picked up _JAVA_OPTIONS: -Dawt.useSystemAAFontSettings=gasp
openjdk version "17" 2021-09-14
OpenJDK Runtime Environment (build 17+35-2724)
OpenJDK 64-Bit Server VM (build 17+35-2724, mixed mode, sharing)

To be more precise, I have tracked down this behavior to have changed in Java 13.

A DESCRIPTION OF THE PROBLEM :
The whitespace separator used by java.text.NumberFormat#format(long) has changed in Java 17, for the French locale (and maybe for others locales which use whitespace separators).
In Java 11, this character was: NO-BREAK SPACE (U+00A0)
In Java 17, this character is: 'NARROW NO-BREAK SPACE' (U+202F)

We have numerous unit tests which rely on this behavior, and which expect the first character.
Granted, this separator character was never officially specified by the method documentation. However:
1) there does not appear to be a rationale for modifying the previous behavior, and we assume the change to be accidental?
2) from what I can see, the second character requires for the whole String to be internally encoded as UTF-16 (two bytes per character), rather than ISO-8859-1/Latin-1, as possible with the former character. So the new behavior doubles the memory used, for all Strings formatted according to the French locale.

REGRESSION : Last worked in version 11.0.12

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Run the test case below

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
I expect the test case below to succeed, independently from the Java version used
ACTUAL -
The test succeeds with Java 11, 12,
It fails with Java 13, 14, 17, with this puzzling message: org.junit.ComparisonFailure:
Expected :1 000
Actual :1 000


---------- BEGIN SOURCE ----------
import java.text.*;
import java.util.*;

import org.junit.*;

/**
 * Succeeds with Java 11, 12,
 * Fails with Java 13, 14, 17
 */
public final class FrNumberFormatTest
{
/**
* This was the number separator in Java 8, 11 and 12.
*/
private static final String NBSP = "\u00A0";

/**
* This appears to be the number separator in Java 13 to 17.
* https://en.wikipedia.org/wiki/Template:Narrow_no-break_space
*/
private static final String NNBSP = "\u202F";

@Test
public void checkNumberFormat()
{
final String formatted = this.formatFr(1000);
Assert.assertEquals("1 000", formatted);
}

@Test
public void checkSeparator()
{
final String formatted = this.formatFr(1000);
Assert.assertEquals(NBSP, formatted.substring(1, 2));
}

private String formatFr(int number) {
final NumberFormat format = NumberFormat.getInstance(Locale.FRANCE);
return format.format(number);
}
}

---------- END SOURCE ----------

FREQUENCY : always