I'm no expert in this but since no one has answered this yet, some limitations would be:
1. Any limitations in the switch they are connected to (e.g. it could be limited to 1 or any number of IP addresses on the port).\
2. Number of devices allowed in the subnet,
3. I imagine the bandwidth (e.g. 100MB/s) would be distributed among the devices.
2 and 3 may mean fairly big numbers are allowed, if there are no other limitations.
Not a HW guy but my understanding is that effectively there is a 3 port network switch embedded in the dual-port ethernet devices. One port is hard wired to the meter itself and the other two are the external ports you can see.
This means only the standard limitations from ethernet itself will apply - 100m wiring length limit and the bandwidth is shared across all the devices on that network segment.
My main concern is based on the fact that most of these meters draw their control power from the metering point and don't have an external UPS backed control power. If any one of the meter losses it's control power due a blown fuse or an outage, all the meters downstream will loss the network connection at the same time. The daisy chain topology may be fine for all devices within one gear/enclosure, but I would plan this carefully if the devices are scattered in different locations.
One would expect the daisy-chain to follow normal Ethernet rules such as 100 meters length? Typically these will be much closer than this in a normal application.
The main concern that I have experienced with Ethernet daisy-chain is when a device in the chain is powered down for service or goes nonfunctional the "downside" of the daisy-chain will be lost. This is not always true with RS485.
One RS485 can "grab" the line and kill all coms on the string but turning off something in the middle will not usually effect anything else on the wire.
I am not totally sold on the Ethernet daisy-chain concept in all application due the real potential loss off data past a device that has gone bad or is powered off.