dunno, but I'd have thought that if you want to test performance, the best measure would be performance at the user end.

for example, you could take the average refresh rates of the half-dozen most reliable API applications (which BF could do because each app logs on with its own its unique softwareId at the start of a session), and then if the average r/r for a sample of the application instances  drops below a critical threshold, then BF would know they have a problem and a big red light would start flashing at BF Towers ... they'd also know which call e.g to update markets or update bets or whatever was having the problem.

i charge $1000 per hour in cash, in advance, and am available outside pub opening times