Error in if (logLike > maxlogLike) { :    missing value where TRUE/FALSE needed

I imagine there's some kind of overflow issue happening here - the model runs for quite a while before this crops up at somewhat random iterations (once at 1500, once at 1100, with same input). Is there maybe a way to robustify the chain and attempt another sample if this occurs?