Wrong likelihood value reported and bounds exceeded

[bthe]

Tested paramin on a simple case and I noticed two things:

• The reported likelihood value was extremely high, and not the same as the comparable single threaded gadget
• Bounds were exceeded on number occasions