As I recall pseudoloc is just randomly picking individual characters to substitute that look like the Latin letters to keep it readable for testing, so it would be something like рг (Cyrillic) ο (Greek)... etc, and can change from run to run. It would also artificially pad or shorten terms to catch cases where the (usually German) term would be much longer or a (usually CJK) term would be much shorter and screw up alignment or breaks.