That's normal. When it completely ices over, start worrying.
I had a 2005 GE SXS that had a couple of those glass tube defroster failures plus a control board failure, several ice cube auger replacements (freakin plastic auger??!?!), water valve failure, and water filter head unit. I spent way too much time pulling everything out and defrosting those coils with a hair blow dryer.
I'll never understand why GE put glass tube defrosters in their fridges. Just stupid. A good quality calrod heater would be ideal.
Speaking of control board another thing I want to see is side by side fridges going back to a plastic damper and EM defrost timer.
........................................................................
On thing which I have theorized about is a horse shoe shaped evaporator with a 100 watt calrod heating element intimately in parallel with the evporator and electrically in parallel with cold control thermostat coupled with an over sized sealed system.
In theory every time the thermostat is satisfied the serpentine heater would come on gradually melting the ice during the off period from the evaporator until the thermostat closes again.
GE did this very successfully in the fresh food section of their manual top mounted freezer defrost refrigerators.
Although two potential down sides to doing it that way in the freezer section could pop: If the door was repeatedly opened the compressor run time would go way up and off time way down- to the point the evaporator could become excessively frosted. Once excessively frosted, the thermostat may never open causing the fridge to run indefinitely and frost to keep building up to the point the unit can't cool.
Another down side would be if the compressor switched on while the frost was only partially melted would result in the water droplets freezing leading to the evaporator gradually building up ice chunks.
I am sure if it was that successful it would have been done by now- elegant to eliminate a possible failure point- but I guess the defrost timer, high wattage heater and defrost termination stat might be better after all.
I'd still like to try it out in a lab if I could to see just how practical the idea could be.