Minor Description
In my case I have CharSequence that implements a "lazy" string which is stored on disk, and although toString() implementation is valid, it is very expensive plus can potentially cause OOM.
Thus CharSequenceUtils.regionMatches() should really do char-by-char comparison, leaving the optimization to underlying CharSequence implementation.
Maybe CharSequenceUtils.regionMatches() could check that passed CharSequence is standard implementation (like StringBuilder, StringBuffer) that has "effective" toString() implementation, but this implementation ends up with creating new String object and thus duplicating the character buffer. So we have classical speed/memory trade-off.
P.S. Line 192 of CharSequenceUtils() reads
TODO: Implement rather than convert to String