The slave may not ACK it's slave address until it's ready, but that should not cause the driver to get to "locked" state.
the protocol to read register from the chip is -- start, send slave addr+ write cmd, send reg addr, start send slave addr+read cmd, read 1 byte, stop.
What i see on the scope is following :
start condition, proper address, nak. (there's no stop!)
then 10 ms later i see another 9 clocks with register address... it's like driver didn't stopped/reset the dma.
any other attempts to access the driver, even with different addresses doesn't produce single clock -- the driver is locked until reset.