Question about Madgwick IMU-Filter


i would like to use the open Source Madgwick-IMU-Filter (Open source IMU and AHRS algorithms – x-io Technologies). In the Madgwick_Report I read "To ensure the optimal fusion of SEqω, t and SEq∇, t assuming the convergence rate of SEq∇ governed by α is equal or greater than the physical rate of change of orientation." and in a forum "The Madgwick IMU filter requires 4-5 iterations on the filter solution for every new data sample."
To my project:
-I use an IMU (MPU6050) and a 8-bit uC
-Madgwick Filter fusion rate runs with 15Hz (SourceCode Variable "sampleFreq"=16)
-IMU-Sample-Rate: Gyro = 500Hz, ACC = 500Hz
Does this mean that the Madgwick filter fusion does not work properly?
Is the "rate of change of orientation" the IMU-Sample-Rate or the Sample-Rate/Frequency in which the uC get new values from the IMU ​​(this would be synonymous with me 16Hz)?
Via Try and Error, I set the Variable "sampleFreq"=880 in the Madgwick-Filter-SourceCode Parameter and the pitch-, yaw- and rollmeasurement works very well (fast, without fluctuations, exactly). Can you explain me, how this works?

I am looking forward to your feedback.

RTIMUlib for Arduino is open source and state of the art. I had no problem implementing it on a Pro Mini and it works very well.