Open any query. Preferably a cheap one, e. g. this one.
Double-click the Execute button.
In the “in X ms” field, you’ll very briefly see the duration of the first execution, and then an incorrect duration for the second one:
in 1456271541357 ms
This is simply the number of milliseconds since the Epoch (verify with date -d @<duration minus last three digits>). The first execution probably resets the start timestamp to 0 ms when it finishes, and then the second execution calculates the offset from that.
The simplest solution, I suppose, is to deactivate the Execute button until the currently running query terminates (result, timeout, error, …).