Query string of POST /leases/{id}/heartbeat, POST /leases/{id}/wait,
POST /leases/{id}/bind, DELETE /leases/{id}, GET /notices, and
POST /notices/{id}/delivered.
Median, p90 and max of what a job’s runs used: the newest runs that neither overlapped
another lease of their session nor were killed, at most as many as an estimate reads.
Cores and GPU load are each run’s average while its session was busy, RAM and VRAM its
peak, and est_secs its run time.
The reason a verdict gives when ending every lease would still leave too little room, so
only something outside Allot can make the request fit, if anything can.