You can do it both ways, where by you have a dedicated instance for each site that uses PHP or one master PHP that spawns children. In your case you are looking to conserve memory usage so you want to have one master process that spawns children as necessary.

The way you define that is by setting each one of your php-fpm pools to be ondemand

pm = ondemand
pm.max_children = 5
pm.process_idle_timeout = 10s
pm.max_requests = 200

This will given you one process and spawn as necessary. This will also limit the maximum number of children to 5 and will leave each one spawned for 10 seconds before killing it so that it can serve other requests. Each process will also be terminated when they have either been idle for 10 seconds or if they have processed 200 requests and you can change these variables based on your memory consumption.

The drawback to this is that you will have an initial lag when the first process needs to be spawned, but it doesn't seem like you are worried about that in this case.

If you want to host the same blog from multiple subdomains you can just set that up in Nginx config:

server_name blog.example.com support.example.com;