What kind of Cold Boot delay can I expect ?

[VictorMustin]

If I deploy a workflow on Runpod that needs to load a Flux dev, a 500MB lora, a lightweight segmentation model + the encoders and VAE, how long would it take for booting ?
I'm not very proficient with Docker and Runpod : If a Docker image contains the weights, does it have to load them every time it runs or are they kept in memory ? I assume that storing the weights in the container would make it kinda big and thus more expensive?

Thanks for the help :)