Multi-server merged System — camera-to-server pinning, failover control, and provisioning for co-located analytics appliances

We deploy an analytics product as N AI appliances, each running a co-located Nx Mediaserver, all merged into one System. Each appliance's plugin consumes frames over shared memory (same-host only), so each appliance must process only the cameras recorded by its own local mediaserver. The architecture depends on each camera being pinned to its co-located server and never moving. Target scale: 5–20 nodes, 60–240 cameras. Nx version: 6.1.2.

Questions:

1. Camera ownership (highest priority): In a merged multi-server System, how do I deterministically control which mediaserver records/owns a given camera, so I can pin each camera to its co-located server? Is this set manually per-camera, or is there a bulk/scripted/API method?
2. Failover control (highest priority): How do I disable camera failover, or pin cameras, so Nx never moves a camera to a different mediaserver? Please give the exact setting — a camera moving to another server would break our co-location requirement.
3. Provisioning/automation: Can a multi-server System be configured via API or config files for unattended install (server merge, camera assignment, recording settings), or is per-server manual setup in the Desktop client required? We ship an installer and want this scripted.
4. Install order: Recommended setup order for a merged multi-server System — configure each mediaserver fully then merge, or merge as we go? Any gotchas merging servers that are already recording?
5. Storage sizing: With each co-located mediaserver recording its own cameras locally, any per-node storage sizing or config recommendations? Any concern with recording storage on the same host as heavy GPU compute?
6. Recording redundancy: If we keep failover off to preserve co-location, what's the expected approach for recording redundancy — or is local-only per-node recording the standard model for this pattern?
7. Upgrades: How do mediaserver version upgrades work across a merged multi-server System — rolling or all-at-once? Can servers run mixed versions during an upgrade?
8. Health monitoring: Supported way to monitor per-server health across the System (API endpoint), so our coordinator service can detect a down mediaserver and exclude it?