Wireless Sensor Networks (WSNs) comprise embedded sensor nodes that operate autonomously in a multi-hop topology. The challenges are unreliable wireless communications, harsh environment, and limited energy and computation resources. To ensure the desired level of service, it is essential to diagnose performance issues e.g. due to low quality links or energy depletion. This paper presents remote diagnostics and performance analysis that comprise self-diagnostics on embedded sensor nodes, the remote collection of diagnostics, and the design of a diagnostics analysis tool. Unlike the related proposals, our approach allows correcting detected problems by identifying the reasons for misbehavior. The diagnostics is verified with a practical WSN implementation. It has only a small overhead, less than 18 B/min per node in the implementation, allowing the use in bandwidth and energy constrained WSNs.