The benefit of adopting Weibull distribution, compared to the widely adopted exponential one, has been demonstrated for railway signaling applications during ESREL2017. In particular, the use of field-return data allowed increasing the accuracy of preventive maintenance scheduling, based on a Weibull approach. Maintenance scheduling is not the only process that can benefit from this generic approach. This paper presents a methodology to detect safety and reliability weaknesses, of a safety-critical system, by adopting Weibull distribution instead of the exponential one. What if field-return data infirm the hypothesis of constant failure rate that is the common approach adopted within rolling stock field? In railways, safety targets for electronic equipment are defined by EN50129:2003 standard. Generally, it is required that a safety-related system must not present systematic failures against safety, and that random failures are tolerable if their frequency is compatible with the target defined within the CENELEC standard. The quantitative assessment is made on the basis of well-known methodologies such as Petri networks or Fault-Tree Analysis, which are based on failure rate metric. Failure rates are used to statistically model the basic event of these analyses and their value sizes the maximum allowable latency of failures in order to fulfil the safety target for which the system has been designed. Relying on field-return failure data, Weibull parameters have been calculated for an existing electronic signaling system and a comparison with existing predictive reliability data, based on exponential distribution, is provided. The design of a safety-critical system typically relies also on redundancies in order to meet reliability/availability requirements. Equipment or sub-equipment are designed with units in parallel in order to avoid service failures when one of the units fails. The architecture of a railway signaling system must not present ‘single points of failures’ in order to meet reliability targets stipulated within the contract with the customer. The additional redundancies are designed according to the specific requirement to meet and according to the reliability of each of the units. If field-return data do not confirm the hypothesis of exponential distribution, reliability figures might show weak values that could not be corrected even by adding further redundancies to the considered system.