If the UART TX pin defaults to a high value, the pull up will still drive the EOMA.
i think.... i think the pull-up needs to go on the other side of the diode.
through the 68 pins. (Connecting to external 3.3V is no good -
*sing-song* nooOOo, that's not good enough to say that: a solution has to be found because that's what the situation is, end of story.
The second diagram is best: http://www.gplsquared.com/eoma_boot/eoma_boot.html#uart_repair
1. It uses power from MEB 2. It disconnects power to EOMA if the UART is pulled out 3. It can prevent power from pull up resistor reaching EOMA if the UART TX line is driven low. 4. It prevent power from PC reaching the EOMA through the UART TX line. 5. The worst case situation is UART holds TX line high and power reaches the EOMA through the pull up resistor. The EOMA may not power cycle properly if that happens.
But it is unlikely to get damaged because the 1.8K / 3.3K resistors places a limit on how high the voltage on the pin can rise. It can rise to about 3.3V and no more. The maximum current drain by CPU is about 2mA which is not enough to do damage (as far as I know).
All other solutions seen so far (other than connecting to 3.3V internal CPU power inside EOMA) is flawed because enough power can reach the CPU through the pull up resistor to keep it from shutting down properly. Some combinations will not allow the CPU to reset until the MEB power is cycled even if the UART is not connected. That is a dopey solution.