`/v3/models` lists every model twice when `group_name` and `--idle_unload_timeout_seconds` are combined

[hansy2]

With the same config.json as above (one graph tiny in group g):

Setup	GET /v3/models ids
no group_name	["tiny"]
group_name, no idle timeout	["tiny"]
group_name + --idle_unload_timeout_seconds 600	["tiny", "tiny"]

The duplicate stays after the graph has woken up. /v1/config and the log ("built 1 groups from 1 servables") show only one servable, and requests work normally. OpenAI-compatible clients that populate a model picker from /v3/models show each model twice.

Same versions as above (OVMS 2026.4.0.869b2186a).