Specialized feature request for one bot per device, each bot in a docker image.

In this case, each individual bot has its own isolated cache, which is wasteful and cause issues since the bot thinks it 'owns' the partition containing the bot. The current workaround is to limit the maximum isolated cache but still, that doesn't with named caches.

This FR is to find a way to share data across each docker container, without inducing corruption. This would likely mean copying from the shared cache but that's still significantly faster than fetching over the network.