Consider Implementing a Lease Reservation Queue

Systems don't have uniform cost and some kernel issues are system sensitive. What are an agent's current options when a system of sufficient type/size/resources is not currently available for immediate lease?

Consider implementing a reservation queue that would allow an agent to request access to a currently busy system resource when it becomes available. Need to consider how an agent is notified when a system becomes ready. How long is the agent given to claim a system once it becomes available?

Idle systems are expensive in a different way. If an agent doesn't respond quickly enough to a lease offer, should the offer move to the next agent in the queue? Does the first agent maintain its place in the queue when skipped? Can an agent observe the existing reservation queue depth?

Finally, what about a priority designation? Should some agents receive priority access to systems? Based on associated project? Overall system utilization may be improved by allowing out-of-project agents to use idle resource. What about breaking a lease when a higher priority agent requests a lease? Should it be supported? How is an agent with a current lease notified of a broken lease? Does that agent with a broken lease receive priority for re-acquiring a new lease?

No files, tests, or entry points are identified. Start by mapping the existing lease and resource-allocation flow, then resolve queue ordering, notification, offer expiry, priority, and lease-breaking semantics; done requires an agreed design precise enough to implement and test.

Written by the indexing model from the issue text.