Simulated annealing may (rarely) print out SA trace crash for seqname. I'll work around it, don't worry. This is a known problem. It has to do with numerical precision problems inherent in the simulated annealing algorithm. If the package detects an error of this type, it punts, and does a Viterbi alignment of that sequence instead. The warning is for information only, reminding me that despite my best efforts, there are still numerical instabilities in the annealing code. You can ignore it safely, your results will be fine.