<p>i managed to get it to over 420 KB/s (that's 3 Mbps of effective bandwidth--useful data transferred, the JTAG is running much faster--without using fast data channels!) by committing a few crimes in the implementation</p><p>at this point the debug probe is so fast it's outrunning the serial flash by a silly margin, and i need better handshaking to not read garbage from slow peripherals</p>