Correct, persistent headless client doesn't perform in the way you'd think.
Persistent AI calculations is actually a parameter in the server config file, set to 0 or 1.
When persistent=1, the server is always running the loaded map and AI run around all the time.
When persistent=0, the server is always running but the map unloads and the AI offload until a player comes online. (this is our preferred setup)
Some information for our readers:
Headless Dedicated Server and Headless Client are almost identical in setup and usage but the misunderstood part is what they actually do.
The Headless Dedicated Server (HDS) hosts the map, host transactions between clients/server, and CAN compute AI calculations by itself.
Headless Client on the other hand acts as a client, using -client, that connects to the designated server to take over AI calculations, this is ONLY used to distribute the workload over multiple processors or cores depending on the usage.
A headless client is only used when two conditions need to be met:
When the HDS is having resource dependency issues while processing AI calculations. (ie, the HDS is maxing out on CPU usage as an application)
and when hardware configuration allows for this style of setup. (ie, either a very powerful box or multiple boxes on the same LAN)
I hope this helps in explaining the persistent parameter as well as the similarities and differences between Headless Dedicated Servers (HDS) and Headless Clients.