For more than 50yrs, RC servos have been vulnerable to localized gear-tooth wear and breakage. A key Bittle design advantage is the spring mechanism in the limbs that cushions gears from gear-tooth shock loads.
Designers typically choose metal gears as needed, but with added weight and cost compared to plastic gears. For multi-DOF robots with the same servo along a limb, an easy cost-structure opimization is to specify metal gears for servos near the body, subject to greater lever forces, and specify plastic gears for servos farther outboard, where forces are less. Mass is also best concentrated near center-of-mass for vehicles and mobile robots.
For some designs, periodicly rotating servos may extend service life, by better distributing wear on gear teeth. Ultimately, RC-format servos are not most-optimal for long heavy-duty service life of locomotion and other extreme load-bearing use.
I personally feel that how to select a servo mainly depends on the load that the robot has to bear.
I agree that the wear is an issue. I have been looking into alternative options. Like all the hobby work done with brushless motors and FOC control paired with Hypoid gearing. But fitting it into submicro servo spaces. And, making it rectangular. I think the biggest game changer to RC servo market would be pushing an ISO like standard for outer size geometry. And manufacturing the servos as more modular components. You can pick your gear material, reduction and DOF, bolt it down to your brushed, brushless, or coreless motor module and select your potentiometer, or encoder to feedback your analog, digital, or smart driver.
I disagree with plastic servos further out though. Specifically with quadrupeds, the ends of the legs can encounter situations that require pushing the entire mass of the robot with the smallest fulcrum. Like the difference between doing a pushup with just your wrist and fingers, or with your chest.