Geoff,
this fix has been around for about five years. The replacement frog does require the switch to be removed to access the retaining screws. the end effect is the same as modifying the frog with switch insitu. With a large fine or second cut file, gently file across the track in the vicinity of the plastic frog. The intention is to reduce the height of the frog back to adjacent railhead height. Some recommend small washers under the adjacent rail ends to raise the rail to frog height. To me this is putting the cart before the horse. Fix the problem first by bringing the plastic rail height equal to railhead.
Once accomplished, then using a small flat file, the same width as the flangeway, gently run the file through the flangeway with the file held at an approximate 45 degrees to the base of the flangeway. This has the effect of using the tip of the file as a chisel point and smoothly removed material through the plastic frog. It is personal preference to have a flange bearing frog or not, depending on how deep you recess the flangeway. By having the wheel flanges running on the base of the flangeway, then one must ensure that there is no intermix of manufacturer's wheelsets as makers such as B'mann and LGB have deep flanges and so would ride differently through the frog.
Of course, prior installation of the switch when new, one would have tightened all screw connections under the switch, plus if it is your thing, seal the screw heads and terminal wires to prevent corrosion, particularly if in contact with the ground. I also make throwbar covers from brass sheet, that significantly reduce the amount of debris catching between the throwbar and adjacent ties.
I also attach a brass strip (around 0.030" thick) to the divergent guardrail to better direct the wheelsets through the frog to prevent the flange on the wheel directly hitting the tip of the frog. The strip is simply bent through 180 degrees at the ends of the guardrail and attaches through slight pressure to the plastic guardrail. At times I need to remove the strip as some manufacturer's equipment has too narrow wheelset back to back measurements and so the wheeel seizes in the narrowed gap through the guardrail and adjacent railhead.
There is no easy no work required fix with these switches. Total effort expended around 20 - 30 minutes per switch to be rewarded with a reliable switch that reduces derailments to zero. I also recess the divergent point rail into its adjacent mainline rail to assist the transition to the divergent track on the outside rail in the curve. The straight through point rail does not require 'checking' into its mating rail. This significantly reduces the tendency for a loco to travel straight through a divergent selected rail with resultant derailment.
I have around forty (40) wideradius electric switches (with ultra reliable LGB switch motors) with stainless rail and am pleased with their aftermarket modified performance. A little birdie told me that a modified (from factory) switch is in the future, that incorporates a lot of the mods we modellers carry out.