Some encodings define REPLACE_CHAR, almost always to '\ufffd',
but then go on and hard-code '\ufffd' in the source code anyways.
Here are some:

./ext/DBCS_IBM_ASCII_Decoder.java:82: if (outputChar == '\uFFFD')
./ext/DBCS_IBM_EBCDIC_Decoder.java:116: if (outputChar == '\uFFFD')

and others.

(Of course, it is possible that '\ufffd' is used both as REPLACE_CHAR
and as an internal special character in the implementation. If so,
that is a deep source of confusion)