Certainly, reversed flow can easily contribute to instability; ideally we would position the outlet suitably downstream to a location where we do not observe reversed flow (i.e. achieving fully developed flow). The pathway (and angular momentum change) through the device is likely to establish eddies which will take time to dissipate. You will often see advice on extending the outlet by x-lengthscales, e.g. 5-10 diameters, but these are only guidelines; you may find you need more or less. It may also be fundamentally transient, but at such low-Re I would hope that viscous effects will help to stabilise the flow. Instabilities (transient-like) can also be triggered by numerical effects, e.g. mesh resolution or quality upstream.
But fundamentally determine if you're satisfied with the overall mass balance and whether the flow is suitably stable (with iterations) in the areas of interest where you are investigating behaviour - that's the essence of the advice from Rob; if the instabilities are primarily downstream of your area of interest they may not be influencing your result.
You are navigating away from the AIS Discovery experience