Mainnet: VM deployment never becomes ready on healthy light node 8072 (both micro and full VM; contract rolled back after 10 minutes)

[mik-tf]

Network: Mainnet Client: official dashboard (dashboard.grid.tf) Date: 2026-06-14

Summary: On node 8072, both a Micro VM and a Full VM deployment fail the same way: the dashboard reports "failed to be ready after 10 minutes" and then cancels the contract, so the deployment is rolled back. The node is up and healthy.

Node 8072 (from Grid Proxy):

• farm LiriaFarm (farm_id 3997), status: up, healthy: true, recently updated
• features: zmachine-light, network-light, zdb, zmount, volume, qsfs, zlogs, mycelium, wireguard (light generation; no classic zmachine/network)

Steps to reproduce:

1. Mainnet, dashboard.grid.tf, Deploy > Virtual Machines
2. Select node 8072
3. Deploy a Micro Virtual Machine, or a Full Virtual Machine

Result (same for both):

• Micro VM: "Deployment with contract_id: 2097842 failed to be ready after 10 minutes" -> contract 2097842 rolled back (on-chain state now Deleted).
• Full VM: "Deployment with contract_id: 2097841 failed to be ready after 10 minutes" -> contract 2097841 rolled back (on-chain state now Deleted).

Both deploy types end in the same readiness timeout and rollback, so this looks like the deployment not becoming ready on the node rather than a per-VM-type problem (for the micro case, the rolled-back contract's workload was the network/network-light piece). 8072 appears to be one of very few up+healthy light nodes on mainnet, so we cannot deploy any VM on it.

Questions:

• Is deployment (in particular network-light) currently functional on mainnet light nodes?
• On a readiness-poll timeout the dashboard cancels the contract, which destroys the evidence and makes a slow-but-converging deploy indistinguishable from a genuinely broken one. Could the dashboard avoid rolling back on a readiness-poll timeout, or extend the budget, so a slow deploy is not destroyed before it converges?