MicroSDXC Card not responding to SD Interface commands in 3.3V mode

Posting this answer as an excerpt of the comments, since the issue seems to have been resolved.

The registers in the SD card can be more tolerant to voltage drops than actual memory access, so this could be a power issue, perhaps you can check with your scope that the voltage supply to the card stays within 2.7 to 3.6 V the whole time. – apalopohapa

@apalopohapa Success, Looking at the input voltage, it was fine up until the ACMD41 init, then dropped down to 2. My boss pointed out a diode on a different page of the schematic that should have been tied low. Now I am reading everything correctly, and can continue. I'm just frustrated because time and time again I've had to tell myself to check power when having problems, and I can't just assume it's not an issue because it works half the time. I appreciate all your guys' help. – farraman

Turns out the answer had to do with the way I was implementing the SPI transactions using the Nordic SDK.

My rcvr_mmc function looked like this:

static
void rcvr_mmc (
    BYTE *buff, /* Pointer to read buffer */
    UINT bc     /* Number of bytes to receive */
)
{
    nrf_drv_spi_transfer(&spi, NULL, 0, buff, bc);
}

I assumed my rcvr_mmc was working since I was correctly parsing command responses. However, sector reads weren't working at all. I didn't even see a full 512*8 clock cycles.

Once I realized that was happened, I looked more closely at the Nordic SDK SPI driver and discovered this:

https://devzone.nordicsemi.com/question/66434/nrf_drv_spi_transfer-larger-transfer-size/

Since sector reads pass bc=512, this resulted in nothing really being read. My very inefficient hack for now to get this work is:

int i;
uint8_t small_buf[1];
for (i = 0; i < bc; i++) {
    nrf_drv_spi_transfer(&spi, NULL, 0, small_buf, 1);
    buff[i] = small_buf[0];
}