What happened?

While testing kata with cri-o 1.25, we found that cri-o was crashing sometimes.
Further investigation shown that this is happening when the kata shim returns an error to cri-o on any calls it makes.

This is the same bug as #3991, but this time it occurs only on errors, not on every call.

Looking into the code, it seems the error was introduced by this commit: 40df9c9

It is removing the "replace" directive we had in go.mod to downgrade genproto to a version that doesn't trigger this issue in gogo/protobuf.
The issue was not found immediately because it's not systematic anymore - I'm not sure why.

Downgrading genproto again seems not possible, as opentelemetry requires a higher version of genproto.

What did you expect to happen?

cri-o should not crash

How can we reproduce it (as minimally and precisely as possible)?

Modify the kata configuration to make the hypervisor path point to a non-existing binary.
When starting a container with the kata runtime, the kata shim will error out because of the wrong path.

Anything else we need to know?

No response

CRI-O and Kubernetes version

OS version

Additional environment details (AWS, VirtualBox, physical, etc.)