The article presents the results of an experimental study of a prototype cloud service for synchronizing local states of distributed application nodes. The service under study is an alternative way to manage HTC computing in an enterprise desktop grid environment, based on the construction of a universal cloud platform. The service allows you to aggregate available idle computing resources in one application and reduce the cost of computing. For the study, we built a program that generates the A022008 OEIS sequence of the first numbers of a constellation of six compactly located prime numbers. A method for constructing a standard application that solves resource-intensive search and enumeration problems and operates under the control of the designed service is described. In simulation experiments and load tests of the A022008 sequence generator, the limits of scalability and fault tolerance of applications running under the service were investigated according to typical parameters of network delays and performance of computing nodes.