I was thinking about this feature's interaction with Multiple scheduler enhancement that is currently in progress.
The attribute to specify which resource to be released while suspending a job (restrict_res_to_release_on_suspend) is currently a server attribute. Now in case of multiple schedulers, I don't know if there is any use case to have each of the scheduler have their own list of resources that can be released during suspension.
I'm thinking that if there is any such need then, each of the scheduler can have it's own set of "restrict_res_to_release_on_suspend" as part of their policy and override server's attribute. If it is not present in the policy then server's value will take effect (similar to how backfill_depth is implemented in server/queue).