@MarcLaf, Most of your question here should be asked in the ESXi/Networking Community as these aspects are not really vSAN-related at all.
But with regard to the vSAN aspect and the not wanting "vSAN to freak out" - the cluster won't care (or even notice really) a node leaving the vSphere cluster and coming back with a different Management IP address, however the action of it being removed from vSphere inventory may/will cause the node to be evicted from the cluster (because vCenter told the other nodes that ESXi-01 is gone now).
So you should use MM if possible when changing each node, if for whatever reason this isn't possible (e.g. lack of compute resources or god-forbid....FTT=0 data...) then you can prevent this from being possible by disabling unicastagent updates on all nodes before starting the overall change (and of course reverting it after all are changed).
Check current configuration:
# esxcfg-advcfg -g /VSAN/IgnoreClusterMemberListupdates
Enable:
# esxcfg-advcfg -s 1 /VSAN/IgnoreClusterMemberListupdates
Disable:
# esxcfg-advcfg -s 0 /VSAN/IgnoreClusterMemberListupdates