I don't need to say it because I am sure you already know.
To find out, you will have to isolate changes and test one at a time.
That means, load a larger batch for the next trip and randomize the rounds in case there is a trend from front to back within the batch. That way the differences within a batch are as evenly distributed as you can make them.
Split that batch into two to three session worth. If you can shoot these sessions within two to three consecutive days it would be the best. If you can't shoot them soon enough between sessions, then split one day into two sessions by cleaning in between the two halves. Take your data on every shot.
We expect to learn two things per session:1) are more fouling rounds required before taking data, and 2) what are the stats for speed telling us?
Clean in the same way you do between the previous sessions.
Then... Shoot again.
If the ammo is all from the same batch, the effect of cycles on neck tension and case volume, charge weight errors, etc., are minimized.
If cleaning is the cause, you will see a strong effect again. If there was a problem with charge weights, you will not see those averages shift since you blended and randomized the whole batch.
In seeing your averages shift, is there a change in the SD and ES? Remember, you need about 15 rounds in the statistic to get some confidence in the SD and ES.
The challenge is that sometimes neck tension and bonding can cause such huge variations that you can't ignore it. Storage time, case cleaning, firing cycles, etc., can all have an effect.
When we are on a narrow node, all of this matters.
When you are done with that first test of two closely spaced sessions, if you have a third of that same batch, I would also test them after several weeks. This tells us if bonding plays a role.