Hello assanemd,
what Andre meant are the following scenarios:
1. Why use a single LUN. Create a separate LUN and then create another datastore on this new lun. So now you will have two separate datastores. You can move some of the VMs in the new Datastore. Seems best option for me. As now you will have two datastore and instead of every IO going to a single LUN will be divided between LUN's depending on the VMs on each LUN. Also since these are mission critical VM's it would be suggested to distribute them over LUNs (so that you do not put all the eggs in the same busket).
2. You can increase the current datastore size by increasing the Storage LUN size directly from storage and then by increasing the Datastore itself. In this case also there is no need for extra extent (you simply extend the size of the current LUN at backend storage)
3. You create another LUN and then extend the size of the current Datastore by using this extra LUN as another extent.
With VMFS 5 there is no need to use the 3rd option.
Though it depends on the environment and particular requirement but suggestion would be as above with Highest preference given to 1st options then 2nd and 3rd likewise.
But as I said it depends.
Also on another thought what is running on those VMs? You mentioned these are mission critical systems with thin provisioned disks? If these are mission critical systems with thin provisioned disks then I suppose IO is not major consideration or bottleneck here. Is it?