But I don’t understand why this would change anything, all ports on the RB5009 are connected to the 88E6393X switch?
The 88E6393X is a 11port switch chip: 3x 10GB raw MI and 8x 1GB with integrated PHYs.
One 10GB goes to the CPU, one to the SFP+ and the 3rd is connected to a 2.5GB PHY chip (QCA8081) going to ether1. Ether2-ether8 are connected to 7 of the 8 1GB ports. The 8th 1GB port is unused and hidden in ROS (and also missing in the RB5009 block diagram).
This is also why the ether1 activity/Link LEDs behave slightly different compared to ether2-ether8, as LEDs are driven by the PHYs.
On RB5009 it is possible to see differences in link stability on port ether1 vs. ports ether2-ether8