I've discovered a bug while writing a parser for JSON thread data.

E.g. take https://endchan.org/ausneets/res/1018751.json thread.
Can you tell why ">>1017591" post reference in 'markdown' item is transformed as "&gt&gt1017591" with a broken HTML entity (missing a closing semicolon)?

And why it's even called 'markdown' when it's actually HTML?