That is good news for Grok team. However, most of the time cost comparing to the result is secondary, and better results and conclusions can come from mixing AI brains together.
Having a crowded and sloppy environment, is not a good way to go. When you log in to your account, you should not source a bunch of environment files by default. It is always a good practice to have a simple environment, just whatever you needed be sourced. If you need extra, just source them before using them. Keep your home clean.
I have some experience with different databases and system designs. In one of my last projects I used SQLite. However, I used only one dedicated process for writing and other processes for read only. This might help.
This is the right practice. Why put the burden on the server? It is the job of client to remember, not the server. Server is there to serve requests, not do the remembering. That is how http worked from the beginning and that is why it has been successful.
This is the same old good way from database world. A db client never touches the disk, but sends requests to a server who owns it. Same in your product. Keys are separated from agents and if needed, agents must send request to a proxy who owns them. Old proven design. Good luck.
You can use old school way. You can have a tape backup, ssd, or a device that is not connected to the network constantly. Copy there and unplug the device.
The backups got wiped together with the systems, so they were reachable from same network. A backup the attacker can reach is not a backup. Good they had an offline copy, but a system this important should have that as regular schedule, not depend on luck.