MaxScale has a mechanism for detecting if a routing thread has become blocked, which eventually will cause the watchdog to kill MaxScale.

In a few places a so called watchdog workaround is used when it is known that some activity may occasionally take so long as to otherwise trigger the watchdog. However, in such cases the thread really is blocked, which means that all clients served by the same thread will be blocked, and the REST-API will become blocked as well (when a REST-API call that requires communication with all routing threads is made).

Those places should be rewritten so that the thread is not blocked and that the watchdog workaround is not needed.