Still I don't understand why Encode::decode('utf8' does not fail when reading wide characters in the first round only in the second round.

There was no wide characters in the first round of decoding. You have the string "\320\245\320\260\321\207\320\260\321\202\321\203\321\200 \320\241\321\203\320\263\321\217\320\275", and it only contains characters with values in [0..255]. A wide character is a character with value greater than 255.