Multiple points of failure!
If any process fails – how is it different from using resource? (So processes have to respond even if they are using the resource)
A list of group membership must be maintained by all processes
All processors are expected to be capable of handling requests
So why not just use the Centralised Algorithm?