Your question is pretty loaded with opportunities for exploration that could provide a better and perhaps even more actionable answer, but in lieu of such info I will still take a swing.
You have to look at the machines that you are seeking to decommission, see what's running on them and what is stored on them, and those you still desire, you will need to re-locate to a server you are going to keep. Repeat that with all found services tasks and storage until you have moved it or stopped it. Then turn off the servers for a period of time and make sure everything works as expected. Then remove them and reference of them from EV.
I would suggest that since you are clustered, it complicates things a wee bit but has the added value that you likely have no VSPs or Index volumes local to the machines you are seeking to decommission. You will likely have some cluster services reconfig too.
This is as actionable of advice I can give currently. If it is not helpful to you, I would suggest you perhaps consult someone who can properly assess your EV environment and guide you down the proper path, as each environment is unique and yours isn't small, the path to consolidation will require assessment to build a cookbook specific to your implementation.
Out of interest, what are you leaving behind? Journal data? Are you still actively archiving?
