Main downside is you will require one LCR per circuit (although you can get away with only a 1 and 24 for quite a while).
This setup has a few notable advantages - using fluid AE means we don't need a bunch of whitelist filters, which kill scalability. Not having to deal with cells is also super helpful (letting us store much more fluid with less complexity) and this setup is easy to deploy, simple, as well as applicable to many other multiblocks. Because the LCR will empty the hatches/busses as soon as it starts a recipe, you won't have any issue with downtime unless you're running really fast recipes. Finally, this design allows up to 5 interfaces with patterns per subnet interface, allowing huge pattern capacity.
The basic idea is we're going to use an AE2 subnet to push recipes into the LCR hatches/busses, while ensuring we have proper blocking to prevent recipe conflicts when we're crafting many things at once.
The subnet should have a dual interface (block form) which uses storage busses to connect to the hatches/busses. Notable, you'll need an advanced blocking card on the dual interface, with blocking and loose mode enabled.
Your main net should have up to 5 dual interfaces (thin form) pointed at the subnet dual interface. These should have blocking enabled.
From here, make sure you add a circuit to the LCR controller (if needed), and then use the fluid AE recipes which you can put in the thin interfaces on your main network. Use either another subnet, conduits, or GT pipes to get the LCR output back into a dual interface.
Image 1 shows the meat-n-potatos of the setup. Red is your main network, pink is your subnet, and yellow is the quartz fiber connecting the two.
Image 2 shows how to configure the subnet (block form) dual interface.