Abrupt change of roll and pitch angular rates

I am testing my Quadrotor setup with PX4. The following is the log analysis for a mission that I tested on.

Log analysis link: https://logs.px4.io/plot_app?log=529f29a1-c832-4f46-8498-768ab7df35cf

When in altitude mode, although the sticks are centered, the estimation of roll and pitch is non-zero. And this is leading to a change in x and y position when in fact x and y should have been approximately constant.

I understand that in altitude mode only attitude is zeroed when the RP sticks are centered. The above occurs even after centering the sticks.

Additionally, the drift in x and y could have been caused due to wind. But there was NO wind whatsoever (It was just still quite twilight when I tested the mission).

From the raw acceleration plot in the above analysis, I can sense there is a problem with vibration. But maybe there is other reason for the drift other than just vibration? I am clearly not able to narrow down the possibilities.

Please excuse my terminology here, I am probably not suppose to call it drift when in fact x and y positions are changing due to change in combination of roll and pitch.

If you could help me in trying to narrow down the possibilities for this anomaly, it would be really great.