I think it's a phase most programmers go through - I jokingly call it “code graphomania”. You make up some sort of overengineered project, then keep implementing it even though there might exist better alternatives. You'd experiment with various language features, try to cram in some wild design decisions that have no basis in real life patterns, etc. It's hard to explain concisely, but I do observe this in other hackers a lot.
And I think Karpeles was going through that phase at this time. I vividly remember seeing his blogposts on reimplementing an SSH server in PHP, just to show that it's doable. I wouldn't be surprised to see that code and other similar terribad ideas running in production.
I think a large part of this is the 1 Hour rule. Where when things are really bad whatever you can do in 1 Hour that improve things slightly get done. Then Repeat.
People basically just keep chasing local Optima until the unholy mess becomes self-sustaining as real improvement becomes more difficult and you can always look back and say, well at least we have "backups" even if it's just a copy on another disk in the same machine.
And I think Karpeles was going through that phase at this time. I vividly remember seeing his blogposts on reimplementing an SSH server in PHP, just to show that it's doable. I wouldn't be surprised to see that code and other similar terribad ideas running in production.