I don’t think it is about signing. If you take a given kernel, call it the “integrated” kernel, then you can add modules provided those modules are compatible. If you take the original kernel, and the configuration of that kernel, and create a module compiled against that configuration, then the module should insert.
However, if instead of simply adding a module you changed certain features, then it might be that none of the old modules work. Or in some cases, the new module won’t insert in the old kernel.
During module insert modules are looked for at:
/lib/modules/$(uname -r)/kernel
The prefix of this “uname -r” command is the base kernel version, and the suffix is the compile time setting of the string from “CONFIG_LOCALVERSION”. Did you set CONFIG_LOCALVERSION? The default is usually “-tegra”. If your base kernel is version “5.10.104”, then “uname -r” should respond as “5.10.104-tegra”. What do you see when you run command “uname -r”?
Obviously you had to build the base kernel Image (the “integrated” kernel) since you patched it. Did you start with the existing configuration, and then add in your changes for RT, or did you just add RT configurations without first setting up for the existing configuration? Did you install all new modules, or just your RT modules?
Here’s something about matching to an existing kernel:
https://forums.developer.nvidia.com/t/topic/229155/12
(and you might need an entirely new kernel install, not just a matching module)
Here is more about full kernel build and install:
https://forums.developer.nvidia.com/t/topic/193640/11