I bet I know why : we can trick mathematical sofware in various ways,...,but why should we ?
find_root is "numerical" function to get approximate real root of function in interval....when we know ( "theorically") the root exists !!
the function can be implemented in various ways, and for example, if using a Newton tangent algorithm, it needs only the function to be continuous and with derivative, to get, in finite steps, a better approximate as the one given
In
http://www.sagemath.org/doc/reference/numerical/sage/numerical/optimize.html the help doc gives xtol=1.e-12 (default) and the algorithm terminated when new approximate x_n is in range [pi/2 - xtol, pi/2 + xtol] for your function
For me, "find_root_approximate" would have been better name than find_root