Repair techniques are used for reasoning in presence of inconsistencies. Such techniques rely on optimisations to avoid the computation of all repairs while certain applications need the generation of all repairs. In this paper, we show that the problem of all repair computation is not trivial in practice. To account for a scalable solution, we provide an incremental approach for the computation of all repairs when the conflicts have a cardinality of at most three. We empirically study its performance on generated knowledge bases (where the knowledge base generator could be seen as a secondary contribution in itself).