Significant delay in cloudwatch metrics for aurora db instances

We are experiencing a strange delay in CPU Utilization and DB Connections metrics for any new created aurora mysql db instances. Once the instance is created the metrics data is available as usual but within a time the delay is increasing. After a couple of days the delay is already about 2 hours. Engine version: 5.7.mysql_aurora.2.11.1 Instance type: db.r5.xlarge Region: eu-west-1

This is how it looks like comparing to the old instances. The orange line is the new instance.

All other metrics are up to date. CPU Total for enhanced monitoring is also working normally. Due to such delay it is not possible to create an alarm and CPU column does not show value on the database page.

We have tried to create several new instances in different availability zones. The same issue for any of them. Has anyone experienced something like that? What can we do to resolve that?