The first one is because you are not running the agent as root and some things need root to be monitored (like the postfix plugin).
You don't have to run the agent as root, but if you don't you must be prepared for errors like this because the operating system doesn't give access to some things unless you are root (like /etc/postfix). You can remove plugins (or disable them if you already deployed them) that you don't want to use or can't use if you don't want to run as root.
I haven't seen that second one in a long time and never did figure out why its happening. It is usually self-correcting - I've never seen a fatal error occur when that happens. I think its due to a corruption in the communications to the server - but I don't know.
Never saw the third one - need the full stack. I assume some plugin tried to make a JNDI call to a managed resource but failed for some reason. The core agent doesn't do anything with initial contexts, so it must be a plugin.