Each time a server connection fails from the smart router, disconnectFailedHost method is invoked at the ConfigurationManager, and this one creates a new scheduled future at a fixed rate to ping the failing host.

In the case of more than one server failing at the same time, the while loop is duplicating the checking, and when canceling the scheduled task is not taken into account which one applies.

Proposed solution: instead of having a failedHosts list to loop, pass the concrete failedHost to the pingFailedHost method, and store in a concurrent map, the relationship between this failedHost and the scheduleFuture to be canceled when applies.