In the early days of MEMS (when it was called micromachining) one of the first big development efforts was the TI Micromirrors (DLP) for large projectors, which used fairly standard semiconductor materials to actuate torsional mirrors to steer light. By applying electrostatic forces to a large array of mirror pixels with reasonable voltages to reflect a beam +/-20 deg at high frequency TI was able to do PWM of their light source over the array (and color with a wheel).
However, they chose Aluminum as their structural material. Any mechanical engineer is now having a heart attack, because modulating an Al hinge through % strain at MHz frequencies is a recipe for cracking and failure... BUT IT DIDN'T! The EEs who originally chose the process and structures got lucky... very lucky.
Some of this comes down to decent design and selection of materials, but fundamentally at table top size such a design would fail in 100s or 1000s of cycles. Why didn't it fail? Eventually (after production and sales) it was determined that due to the very thin and narrow supporting flexures (to keep forces and costs low) the crystal grain size was typically significantly larger than the hinge.
As a result there were not defect sites for dislocations to create cracks or hardening... effectively they were single crystal Al hinges in the area of high cyclic strain... most of the time except for the odd occasional pixel... but proper etching and annealing improved even that through process control. Why change what isn't broken?