Aurora Serverless v2 scales up ACUs very often for seemingly no reason

Hi, it appears around 3 times each day, for 4-5 hours each time, Aurora (MySQL) rapidly scales up and down.

This is despite incredibly light use of the database. I changed the ACU limit from 1.5 to 1.0 at 5pm yesterday, which is why the effect seems to reduce those final 2 times on the graph.

Back when the ACU limit was 1.5, the logs indicated the buffer pool size was increasing, then 5 min later decreasing again. A pause for around 10 min, then repeat - for around 4 hours. Since lowering ACUs to 1.0, I no longer see logs of buffer pool resizing. So I think this was a symptom rather than a cause, but it may help someone determine what is happening here.

The performance insights show no slow queries, no difference in network activity, fairly consistent CPU load, memory use only changes during the scale ups/downs, and show there is minimal load on the database overall.

Does anyone know what's going on or how I can prevent it? I would prefer to increase my ACU limit for redundancy, but with this issue and many periods of low load, it would be a waste of money.

You can see in this 1 hour graph the particular fingerprint of event is roughly identical each time.