There are a number of replicated strings in the encodings. For example,
CharToByteCp949 and CharToByteCp949C are almost identical (diff them to
see). This representation is bad for footprint on disk and in memory.
(I discovered the redundancy while looking at compression of constant pool
redundancies.)

Other giant byte-for-byte duplications are in ByteToCharCp930 and
CharToByteEUC_KR. In addition to byte-for-byte equal arrays, it would
also make sense to represent arrays that differ only in one element
differently, for example, by testing for the differing index instead
of replicating the array.