SOUNDEX() Algorithm Returns Incorrect Results for Letters With The Same Soundex Code and Vowel Separators

MariaDB's SOUNDEX() algorithm returns incorrect results when used on character sequences which have only one consonant soundex coding separated by vowels and a starting letter with the same coding as the consonant:

Ex. The proper result for SOUNDEX("Popoff") is P110: Popoff-> P01011 -> P110 (since zeros are dropped and side by side same letters are reduced). Instead, your system returns P000, which is the same result as SOUNDEX("P"), which is clearly ridiculous.

I have not seen any proper Soundex() which operates in the way yours currently does.