Sometimes it happens that eventhough UI shows warn "cannot start pipe, previous execution did not finish", it is actually started, but in that case (pipeline extractor dbpedia + t-sparqlUpdate) it may happen that the data units are not properly initialized. (e.g. certain input processed two times)