Context: Extended circumstellar emission has been detected within a few 100 milli-arcsec around > 10% of nearby main sequence stars using near-infrared interferometry. Follow-up observations using other techniques, should they yield similar results or non-detections, can provide strong constraints on the origin of the emission. They can also reveal the variability of the phenomenon. Aims: We aim to demonstrate the persistence of the phenomenon over time scales of a few years and to search for variability of our previously detected excesses. Methods: Using VLTI/PIONIER in H band we have carried out multi-epoch observations of the stars for which a near-infrared excess was previously detected with the same observing technique and instrument. The detection rates and distribution of the excesses from our original survey and the follow-up observations are compared statistically. A search for variability of the excesses in our time series is carried out based on the level of the broadband excesses. Results: In 12 of 16 follow-up observations, an excess is re-detected with a significance of > 2 sigma, and in 7 of 16 follow-up observations significant excess (> 3 sigma) has been re-detected. We statistically demonstrate with very high confidence that the phenomenon persists for the majority of the systems. We also present the first detection of potential variability in two sources. Conclusions: We conclude that the phenomenon responsible for the excesses persists over time scales of a few years for the majority of the systems. However, we also find that variability intrinsic to a target can cause it to have no significant excess at the time of a specific observation.