There isn't a highest age for testing. When you participate in an experiment, you enter your birth year, and we use that to stratify the data into age bands.
Unfortunately, participant panels are not great at having representative populations. It's been a while since we've put a study on Mechanical Turk, but it famously skewed towards young Indian men.
One of the reasons to ask age and gender is to balance towards representative. It helps you detect and correct for imbalances in the participant pool. However, commercial participant panels are bad at certain demographics, particularly at scale. There simply aren't a lot of 70yo women using UserTesting or Cint. If you insist on having statistically significant quantities of responses from older women in every experiment, you'll exhaust those panels disappointingly quickly.