The formal verification of a concurrent system with a model-checker provides the user with a counterexample trace when a property is violated; however the problem diagnosis still remain a complex issue. Diagnosis is made difficult for several reasons: the trace conforms to a structure that is internal to the verification tool and hence hard to exploit, the trace yields low-level information, the trace size can be large. Traces cannot be understood without the Labelled Transition System (LTS) underlying the model-checker exhaustive exploration; we designed KriQL a query language over the LTS graph featuring a blend of set filters and graph-based operations and we implemented some feasibility prototypes. Benchmark results indicate that a hybrid management system using a graph-based database and dedicated data structures should achieve sufficient performances.