Putting all of the large communities on a single instance is just reddit with more steps. It's good that one of the larger Lemmy communities is not also on the largest Lemmy instance. Lemmy.world suffers a lot of outages (in part because it's so centralized), meanwhile this community remains available.
I have my own instance (gadgetro.id), but it isn't set to private. Are you still able to browse other accounts on the fediverse from your instance when set to private?
The lxc container in Proxmox is at around 500 MB of ram usage, 1.35 GB of network traffic, and less than 1% CPU usage. Although I'm the only one using the instance.
An instance is just a way to access Lemmy. Lemmy.world, lemm.ee, sh.itjust.works, feddit.uk, etc - you can use any of these to access Lemmy and interact with, and see, all the same content.
It's a problem because most communities are hosted on lemmy.world at the moment, but if lemmy.world ever dies they'll die with it.
To elaborate further from the other comment, it's a person running a copy of the Lemmy software on their server. I for example am running mine (and seeing this thread) from https://zemmy.cc. Thanks to Federation all of our different servers are able to talk to each other so we can have a shared experience rather than everyone being on one centralized instance managed by one set of administrators (like reddit is).
This provides resilience to the network. If reddit goes down, reddit is down. If lemmy.world goes down, you can still access the content of every community that isn't on lemmy.world, and if other servers were subscribed to the content on a community from lemmy.world you could still see the content from before the server went offline (and it will resync once it's back up).
If we put all of our eggs into a single basket, we have a single point of failure. If all of the major communities go to lemmy.world then lemmy.world is that single point of failure. Doing that is effectively just recreating the same issues we had with reddit but with extra steps. By spreading larger communities across servers we ensure that the outage (or permanent closure) of a single instance doesn't take down half the active communities with it.