给主办方的通报：舵机过热，以及两个不紧急的请求

Hi — a heads-up on a servo incident on our robot last night, and one small ask at the end.

WHAT HAPPENED
During an unattended remote teleoperation session, our left arm's shoulder-lift servo
reached 61 C. Our thermal guard stopped the session at that point (it warns at 50 C and
stops at 60 C, both well below the servo's own 70 C limit) — deliberately, so that we
park the arm rather than have the servo drop torque on its own and let the arm fall.

Root cause was not the servo. The arm had been driven into an extended pose and then
simply held there: 6821 of that session's 6889 control ticks (99%) were "holding
position". Nothing in our software made it stand down, so the joint carried the arm's
weight continuously until it overheated.

WHAT THE DATA SHOWS
Same-pose comparison, both arms parked at home, logged automatically:

    left  shoulder_lift  37 C / load 224 / 12.1 V
    right shoulder_lift  36 C / load 216 / 12.0 V

So there is no hardware asymmetry between the two arms — 1 C apart at the same pose.
The overheating was caused by the pose and by nobody being there, not by a bad part.

Immediately after the 61 C event the servo did misbehave: driving it 2.6 degrees produced
15-33 bus communication failures, while its neighbours on the same bus, at HIGHER load,
had none. We initially read that as a failed servo. It was not. After roughly an hour with
torque disabled we ran a load-gradient sweep and it came back completely clean:

    amplitude   angle    ID2 fails/load   ID3 fails/load   ID4 fails/load
        5       0.4 deg     0 /  32          0 /  36          0 /  37
       10       0.9 deg     0 /  78          0 /  77          0 /  81
       15       1.3 deg     0 / 149          0 / 120          0 / 116
       20       1.8 deg     0 / 195          0 / 148          0 / 143
       25       2.2 deg     0 / 239          0 / 171          0 / 164
       30       2.6 deg     0 / 255          0 / 207          0 / 188
       40       3.5 deg     0 / 291          0 / 223          0 / 220

ID2 is the joint that overheated; ID3/ID4 are healthy references on the same bus, same
power, same wiring. ID2 now sustains a HIGHER load (291) than either reference (223/220)
with zero failures, including at the exact amplitude that failed 15-33 times while it was
still hot internally. Three further repeats at that amplitude: zero failures.

Conclusion: thermal derating, fully recovered. The servo is fine.

WHAT WE ALREADY HAD, AND WHAT WE ADDED
Already in place: a thermal guard that samples every servo round-robin and stops the
session at 60 C, with a re-read confirmation so one corrupted byte cannot stop a session.

Added last night:
  - per-servo LOAD and VOLTAGE are now logged alongside temperature, at zero extra bus
    traffic (the guard cycles through the three registers instead of only reading one).
    This is what let us separate "this joint is working hard" from "this joint is broken";
  - an idle auto-park: after 3 minutes with no operator input the arms retract slowly to
    the home pose and hold there, keeping the gripper state so a held cup is not dropped,
    and keeping torque on so the arm does not fall. Any keypress aborts it immediately.
    Measured basis: at home the same joint sits at 37 C / load 224 and stays there.

THE ASK
Two things, neither urgent:

1. If anyone experienced with these SO-101 arms is around, could they spend ten minutes
   checking the mechanical resistance of our left shoulder-lift joint by hand, with the
   power off, against the right one? A servo was already replaced at that position once
   before. Extra mechanical resistance would make that joint run hotter than its twin for
   the same work, and no amount of software protection fixes that.

2. If a spare Feetech STS/SCS servo happens to be available to keep on the shelf, we
   would appreciate knowing where to find one — not to fit now, but so a failure a few
   days before Demo Day does not become a blocker.

Everything is safe right now: torque disabled on all 17 servos, all of them at 32-36 C.