Is it worth revealing the rest of the schematic? I can see there's ethernet, audio, RGB, etc not shown. Do you have a second identical board to confirm the behaviour is exactly the same? Have you disconnected your USB/programmer? What's the voltage on battery +? 3V3_CORE? EN? Can you confirm that yo...

Schematic?

That would be really inconvenient if the ESP32 restarted after 9 seconds in download mode! That wouldn't be enough time to even flash most binaries. Fortunately (or unfortunately for you) the ROM bootloader disables the watchdog for download mode. See if strong external pullups can ensure correct le...

EXCVADDR (which I understand to be the address resulting in LoadProhibited) is 0x8, which is an invalid address but your NULL checks will have no effect.

A quick search for "CONFIG_FREERTOS_UNICORE" in bootloader_support reveals a lot of initialisation dependent on this, many of them calling into ROM functions. I'd expect to find something like an "assert(chip_efuse_is_dual_core());" somewhere too if I looked hard enough. So it's definitely not offic...

ESP-IDF bug, set it to ON to increment refs first, then you can follow it with OFF.

Pullup is enabled by default on this pin. See datasheet appendix 4 "IO MUX". These are typically around 40-50k iirc.

It sounds like you'll have plenty of flash available for 2 app partitions and that will definitely be the easier and more reliable approach. If the only motivation for updating via the bootloader was reducing the flash footprint of app partitions then this is the way to go. I hope I haven't talked y...

Either implement a basic bootloader SDMMC component from scratch by studying the SDMMC component and TRM, or copy the existing SDMMC component to bootloader_components and go through it replacing all incompatible dependencies (eg. any FreeRTOS stuff, GPIO functions from the driver component, etc) wi...

Try increasing the stack size. 2048 is on the low side.