There's no changing the ESXi behavior, if the device changes how it displays it ID when first in-properly access, I'd argue this is a defect with how Coral has designed their hardware as this is the first time I've ever heard of such a strange behavior and while this maybe observed w/ESXi, it is important to understand why this is occurring. Is it because the ESXi hypervisor attempts to access/probe the device because it has not been setup to be passthrough or is it when VM attempts to access it for first time but the software hasn't been setup, so connection has already been reached?
What I'm more interested in is whether someone has spent the time to try different variations to help isolate where the initial connection is attempted and whether this is happening at the Hypervisor layer or within the VM. Depending on the behavior, there may be some options or at least more information to find a potential solution. I've got a colleague who recently purchased this device for their own testing, so I've shared this thread to see if they're running into any setup issues. I also noticed some of the issues observed is the use of container to access the device, this potentially may add another layer of complexity.
Again, I rather not jump to any particular solution until we better understand what is causing the initial connect and where it is coming from. Happy to chat offline to debug further if you've got an env running latest ESXi 8.0 as any tweaks/changes would get greatest benefit w/our latest release of ESXi
