Magnetic reversals aren't terribly regular so (like volcanoes) they can't be 'overdue'. They seem to follow a broad pattern of a gradual dwindling of the global magnetic field over a few thousand years with the appearance of several local magnetic poles around the globe, followed by a flip and a gradual strengthening of the global field.
The duration of the reversal itself isn't known with huge precision, but anything between 2000 and 12000 years seems to be the best fit; although at least one paper proposes the most recent Bruhnes-Matuyama reversal about 0.781My was complete within 200 years.
Though, just to make it more complex, the apparent duration of the reversal in any particular location is incredibly varied as it relies on issues such as the geomagnetic latitude and local non-dipole components of the Earth's magnetic field during the transition.
God I hated palaeomagnetism when I did my MSc - it's absolutely bloody brilliant - until it isn't. Though it was a damn sight more useful than the radioactive dates I was using which were very much 'pick a number between yesterday and a hundred million years ago' due to hydrothermal contamination. Ooops - I digress.