Scientific workflows composed of independent distributed web services allow a user to easily orchestrate dataflows and processing tasks, and handle interoperations with ever more intricate workflows. Typically, these integrated workflows are complex, tightly-coupled, and computational and data intensive because of the workflow invocation management. One side effect is the increase in no-response states as unpredictable events and infrastructure inconsistencies interrupt invocations and cause failure, which are hard and heavy work to manually debug. We propose middleware for running and debugging integrated workflows called RDW utilising the Taverna Server invocation engine which helps the user investigate complex tightly-coupled workflows. Two debugging modes, sequential and parallel, are provided for inspecting the sub-workflows of these integrated workflows. Also, interactions between the user and the server are orchestrated seamlessly by the RDW which manages workflow invocations and provides running and debugging logs. The execution time of our RDW is slightly less than the standard workbench, although the sequential and parallel RDW debugging modes have slightly different running times. RDW offers stable workflow invocation which can significantly reduce the total execution times. The middleware is freely available from our website (http://bioservices.sci.psu.ac.th).