With the introduction of remote FDB we have a few places where we serialize and deserialize in the same code that is doing some higher level logic.

We need to reevaluate the current implementation and explore if we should introduce an abstraction over the serialization/deserialization code with the goal to hide future protocol changes from the calling code.

See RemoteFDB::RemoteFDB for an example, we construct the bytebuffer and build the message byte by byte directly. IMO the code here operates on very different levels of abstraction and should be separated.