Splitting geographically separated non-backbone OSPF areas?

I'm a network engineer for an air gapped military network. We run a worldwide WAN in which we control the WAN routers, downstream are customer owned Nexus's that hang off our routers which we do not control. Currently we have our own internal OSPF process, we have static routes pointing downstream for the customer owned IP space that we redistribute into OSPF to advertise the sites to each other. They just have a default route pointing to us because we are their only way out.

We've been mandated to integrate them into our OSPF process which is fine, not a big deal. My question for the design, can we incorporate all of them into one non-area 0 area? Or should each site be in its own area? We have geographically separated sites spread out across the US, Europe, and Asia. I'm not sure if it's pertinent but we are going to configure each site as a total stub because we are their only way out and because we also need to advertise a default route down to them. I don't want to use default-information originate because each of our routers have their own individual default route to get out to their respective "internet" and I don't want one site learning another sites default route.

I was leaning towards putting each site in its own area because I don't want every device to have to rerun the SPF algorithm if the downstream link flaps or something but my coworkers want to keep all the customer facing interfaces in the same area for simplicity. Google tells me there is nothing inherently wrong with non-connected non-area 0 areas but I can't seem to find anything definitive especially with high latency wan links.

Shitty drawing: https://i.imgur.com/zHArBMc.jpeg

Thanks in advance for any help/insight.