Hmm, again? I'll try to repro at my desk.

So, I flashed that image onto a servo_v3 on my desk, and although there are errors reported, servod seems to be running. Comments about the errors are below. 

servod -b chell on a servo_v3 image yields the following:

2016-06-06 19:24:47,005 - Servod - INFO - Initialized ec_ec3po_interp_connect to on
2016-06-06 19:24:47,008 - Servod - ERROR - Unable to determine driver for jtag_vref_sel0
2016-06-06 19:24:47,010 - Servod - ERROR - Problem initializing jtag_vref_sel0 -> off :: 'drv' key not found in params dict
2016-06-06 19:24:47,012 - Servod - INFO - Initialized jtag_vref_sel0 to off
2016-06-06 19:24:47,014 - Servod - ERROR - Unable to determine driver for jtag_vref_sel1
2016-06-06 19:24:47,016 - Servod - ERROR - Problem initializing jtag_vref_sel1 -> off :: 'drv' key not found in params dict
2016-06-06 19:24:47,018 - Servod - INFO - Initialized jtag_vref_sel1 to off
2016-06-06 19:24:47,027 - Servod - ERROR - Getting usbpd_ec3po_interp_connect: 'NoneType' object has no attribute 'get_interp_connect'
2016-06-06 19:24:47,029 - Servod - ERROR - Problem initializing usbpd_ec3po_interp_connect -> on :: 'NoneType' object has no attribute 'get_interp_connect'
2016-06-06 19:24:47,031 - Servod - INFO - Initialized usbpd_ec3po_interp_connect to on
2016-06-06 19:24:47,041 - Servod - ERROR - Problem initializing uart4_en -> on :: [Errno 121] Remote I/O error
2016-06-06 19:24:47,043 - Servod - INFO - Initialized uart4_en to on
2016-06-06 19:24:47,055 - Servod - INFO - Initialized uart4_on_flex_en to on
2016-06-06 19:24:47,059 - servod - INFO - Listening on localhost port 9999


The usbpd_ec3po errors can be ignored. That's a known issue and the fix for that landed in 8381.0.0. (https://chromium-review.googlesource.com/#/c/342467/)

So then it seems to be an issue with the jtag_vref_sel* controls. Does servo_v3 have JTAG support? I don't think it does, but I could be wrong. In the servo_v2_r1 overlay, the driver for the jtag_vref_sel* controls is the tca6416 i2c IO expander driver. It's not defined for the servo_v3 overlay.

Considering that servod is running for me, I think that those jtag errors are "normal" since they pertain to the PD MCU UART which is not officially supported on servo v3. Those errors should be silenced as well if that's the case.

Are there any logs from the servo v3s that can't start servod?

Also, regarding the error messages, I would expect chell, lars, and sentry to have the same error messages, as they all inherit from the same overlay, servo_glados_overlay.xml.

Just launched servod -b chell on servo V3 running 8421.0.0 connected to a chell and see similar output as #c3

servo V3 does support JTAG but in this case (glados based designs) its been hijacked to support UART access to the USBPD controller.

The jtag control issue is expected and should be benign as the V3 control, 'jtag_vref' defaults to 'off'

<sigh> I went back to check chell again, and I'm no longer
seeing the servos as down.  I checked the historical logs,
and it seems chell was never a problem...

I re-checked, and there seems to be a problem of some sort
with guado and nyan_kitty, but I can't find evidence that the
problem is/was in hdctools, or even in software, as such.

There's something odd, but I don't have evidence that justifies
code changes, and no time to find the evidence right now.

I'll refile something when/if I have something constructive to
say.