Deterioration of the insulation can occur in a number of different ways depending on the conditions.
In dry conditions, ozone formation can physically oxidise the insulation and leave behind the oxide which will appear as a dust or powder.
In moist conditions, oxides of nitrogen (forming nitric acid with the humid air) can literally eat away insulation as well.
The main catalyst is still the low clearances giving a high electric field stress around certain points of the insulation. It sounds like it may well have started around the silicon rubber inserts, which are oxidising, and a silicate powder being left. The tracking will occur once the pathway becomes polluted enough to sustain a burning or corrosion of the insulation. It might be worthwhile doing a PD test on the switchgear (on the busbar mainly) which should confirm the corona presence. Was PD testing done on installation, if so, were the inception and extinction voltages acceptable? if not, this is a good example of why you don't accept the factory test results as being what is there on site. Many authorities have gone away from commissioning PD tests and opt only for an HV test. The PD results provide a valuable reference for new switchgear (especially epoxy gear on which DDF tests can be close to useless).
My concern is the speed at which the deterioration has taken place. We would regard this sort of deterioration as a long term (10-20 years) issue. The fact that you have reached it in 3 years says that the reaction is being driven by detrimental conditions that are way above normal.