MEMTRACE_FLEET_MEDIATOR=1 has no observable effect on mediator_mode

[Regis-RCR]

Summary

Setting MEMTRACE_FLEET_MEDIATOR=1 in the engine or MCP environment changes
nothing I can observe in mediator_mode. The flag does reach the process
environment, but every fleet surface that exposes a mediator mode still reports
advisory.

Environment

• memtrace 0.6.11
• macOS (Apple Silicon)
• Embedded MemDB, fleet tools called over the MCP transport

Steps to reproduce

1. Start the engine / MCP with MEMTRACE_FLEET_MEDIATOR=1 in the environment.
   (Confirmed present: the value shows up in the running process environment.)
2. Trigger a conflict that raises an escalation (a destructive change over
   another agent's live intent on the same symbol).
3. Inspect the mediator mode wherever it is surfaced:
   • fleet_status reports mediator_mode: "advisory".
   • the escalation's mediation request reports the same advisory mode.
   • fleet_preflight carries no mediator-mode field at all.

Observed

The flag is recognized as an environment variable but yields no observable
mediator_mode delta in 0.6.11 (same-host, embedded). The mode stays advisory
everywhere it is reported.

Expected

One of two things. Either the flag changes mediator behaviour in a way an
operator can see, or, if it is not wired in this build, it is documented as a
no-op (or removed) so nobody relies on it.

Note

This one is minor. The flag may well govern some other path (an auto-resolution
route, for instance) that these tools simply do not surface. If that is the
case, a short note on what it actually controls would close the gap.

[blocksorg]

Mention Blocks like a regular teammate with your question or request:

@blocks can you add technical details to this issue?
@blocks make the following changes ...
@blocks create an issue from what was mentioned in the following comment ...
@blocks explain the following code ...
@blocks are there any security or performance concerns?

Run @blocks /help for more information.

Workspace settings | Disable this message

[Alex793x]

You called it — the flag was read on every call, but the parser only recognized off and auto; everything else (including 1) silently fell through to the advisory default. An exact no-op, as observed.

Fixed in v0.6.17: boolean-style values now map to real modes (1/true/on → auto, 0/false → off), unrecognized values log a warning naming the valid set (off|advisory|auto). Two notes for your setup: the env var must be set in the process that executes the fleet tools (the agent's MCP server, not necessarily the daemon), and what auto changes observably is escalation handling — deterministic-safe conflicts resolve without a human verdict, which you'll see in the escalation's mediation outcome rather than in fleet_preflight.