Custom extensions of the Language Server Protocol

For some languages (concretely thinking about hardware description languages such as VHDL or Verilog, but this probably applies to more), there are certain desirable features that are not covered by the language server protocol, but that are small enough so these features would likely not be implemented into the mainline protocol. As one concrete example: These language description languages usually have a notion of instantiated units that form a hierarchy of sorts. This hierarchy is similar to the notion of a CallHierarchy or TypeHierarchy yet cannot be expressed satisfactorily or semantically correct with either capability. At the same time, it is also not practical to implement support for such InstantiationHierarchy (or similar) capabilities into the mainline protocol because most languages will never use this capability it and therefore most editors won't enable support for the capability.
In my opinion, these kinds of features would best be implemented as part of a custom extension of the Language Server Protocol. Plugins in editors (thinking especially about VSCode, but this is widely applicable) could tune into these extensions and provide the client side implementation of the protocol. The Base LSP is still applicable, but the method names, appropriate parameters and responses would be custom.
My question therefore is: Is there a 'standard' way of extending the Language Server Protocol to implement these niche features? I.e., is there a certain namespace or schema for method names that can or should be used to implement custom capabilities and does LSP even allow or encourage extensions? And if it doesn't so far, would this make sense?

No files, tests, or entry points are named. Begin by reviewing the LSP extension question and its InstantiationHierarchy example; done would mean reaching a documented decision on whether custom capabilities are supported and how they should be represented.