The speedruns website tries to remove illegitimate runs. Those that have been found out won't be in the data.
I think it could be worth plotting the improvements in TAS runs. Would the pattern of diminishing returns be the exact same as the human speedruns? The world records you're tracking are limited by execution, and as new techniques are discovered the difficulty of completing a run increases, so good runs take more attempts, more effort, and time. Many techniques that are discovered and used in TASes are even too precise for human fingers. Human WRs are founded on technological discovery but are chained to execution. The only way to not conflate the two would be to look at TASes, which is where the technological bleeding edge is.
Since anxiety is a crucial link in the cycle, aren't benzodiazepines a relevant therapy?* Why not?
* If a benzodiazepine does address nociplastic pain couldn't taking one just for a few days be used to diagnose nociplastic pain? 'Real' pain would not be addressed the same way by a benzodiazepine. (They do have some analgesic properties but they're no opioids, not by a mile. Normal doses are just anxiety-killing.)