By watching the bluer bits of the image, you can see that it seems that the camera was very quickly (almost immediately it seems?) centred on the cliffs to the bottom right, then recentred a few frames later but at an angle (probably rotated around 10 degrees anticlockwise). The next few frames are the actual explosion, where the offset angle seems to remain consistent, and then, interestingly, the HUD is almost perfectly recentered at 0 where it remains consistently
Through 19s, again the person recording this from the screen seems to either zoom or move the camera very quickly towards the screen, before immediately resetting.
It's clear that this weirdness is being introduced by the person recording the video (not from the drone itself) because the HUD also moves and blurs (you can see what it looks like when the drone camera pans throughout second 23/24. Note the HUD remaining static and clear).
Second 26 is also interesting, because you can see static bright spots suddenly seem to begin moving. However, I think this is actually a rotating lens cover on the drone camera which is introducing these artifacts. You can also see that immediately after this, the HUD display is different. A different number (I believe it's 80, based on a frame at 37s) is displayed, and the centre of the HUD is a crosshair rather than a small circle. I believe this new lens is slightly shorter, and is supposed to give a wider view of the explosion that just happened.
If you advance slowly through second 37-38, you see the camera 'zoom out' from 80 to 402, and it seems both the drone and the person filming this's cameras needed to recalibrate their exposure settings.
It's not quite enough for me to say it's definitely suspicious, but those movements at the 18-19s mark are definitely a bit odd. It could potentially be artifacts of YouTube automatically stabilising the video, but there's definitely something unnatural here.