Yes, This ActiveSense message is not sysex though, which is why the problem seems strange.
I should give some background as to how the Minicommand handles MIDI. Basically there are interrupts listening on the input MIDI ports. Whenever a byte is received the currently executed code is paused, and MIDI interrupt code is executed. This is really light weight code that basically services the MidiClock (updating counters and handling start, stop and continue messages. song pointer etc), for all other messages such as SYSEX the MIDI bytes are stored in a Ring Buffer which is left sitting there waiting to be processed by the main program loop.
In normal operation the Minicommand falls into a main loop where it refreshes the GUI (handles the button presses and updates the LCD), it also handles midi by reading a byte off the Ring Buffer. Depending on what the midi message is this may cause the Minicommand to execute code which will cause the main loop to pause until the code is finished executing.
For MIDI Output the concept is similar. There are interrupts on the output ports. Data to be sent is stored in a Ring Buffer. The Ring buffer is processed one byte at a time. Whenever a byte has been transmitted on the MIDI port an interrupt is triggered which causes the next byte in the buffer to be sent. There are some edge cases where by if the buffer is full, the interrupts are skipped and the entire buffer is drained. For MIDI clock messages, (such as merging the midi clock from the 2nd input port) the buffer is skipped and the data is sent immediately (i.e. when the port is next free). (*The above applies to MIDI Ctrl 0018, 0017 and below incorrectly handled midi output transmission).
When you send big chunks of Midi data such as SYSEX dump, the main loop will actually be halted and all processing resources devoted to encoding and placing the data in the output Ring Buffer. Interrupts along the way will transmit this data, unless the buffer is full in which it will drain.
So getting back to ActiveSense. It’s a MIDI Status message that should be sent after 300ms of inactivity to signal to the connected device that the connection is still alive. In the Minicommand’s case there is a timer interrupt that is decreasing a 300ms counter until it = 0, and then sends out the status command if midi data has not been transmitted during this time. I suspect that the implementation here may be bugged. Why it’s causing the sysex data to be corrupt I have yet to work out.