I'm still seeing this. I just noticed the delcsn.bv_val[delcsn.bv_len] = '\0'; in syncprov.c 1.152. Any chance that might be manifested with these symptoms? (I'll put it in production anyway, looks like a good idea.)