Ekf2 doesn't use data from sensors immediatly

[Slimercorp]

Describe the bug
EKF2 timestamps reports that ekf2 doesn't use mag, baro, gps and other sensors immediately (when topic was updated). EKF2 uses it after 1-2 self iterations. Why?
For example:
0 iteration: IMU was updated (timestamp 4000), mag wasn't updated. EKF2 use IMU data only.
1 iteration: IMU was updated (timestamp 8000), mag was updated (timestamp 6500). EKF2 use IMU data only.
2 iteration: IMU was updated (timestamp 12000), mag not updated (timestamp 6500). EKF2 use IMU and mag data. Relative timestamp of mag is 5500 microsecond or 50 (ekf2 timestmap value)

It may see on relative timestamp. If relative timestmap > 40, then it means ekf2 doesn't use data immediately.

The problem meets in all logs, that i can see here http://logs.px4.io/.

To Reproduce
Use, for example, firmware v1.8.2. Look on mag relative timestamp.

Expected behavior
Perfectly all relative timestamps must be lower 40 (4000 microseconds, it's time of update IMU and ekf2). Other words, new data from mag, baro, gps sensors use immedeatly

Log Files and Screenshots
https://logs.px4.io/plot_app?log=1a0d7681-2c9d-462c-877f-f4338718de56

Is there explanation of this problem?

[priseborough]

EKF state prediction and fusion cycle is running on a delayed time horizon as explained here: https://dev.px4.io/en/tutorials/tuning_the_ecl_ekf.html