If there are many listeners to bucket notifications or to the trace
subsystem, healing fails to work properly since it suspends itself when
the number of concurrent connections is above a certain threshold.
These connections are also continuous and not costly (*no disk access*),
it is okay to just ignore them in waitForLowHTTPReq().