The autopilot behaviour differed considerably from the Boeing documents:
By 05:38:45 the left AoA sensor was reporting (but not displaying) over 74deg AoA, yet the LH AP was engaged successfully some 37 seconds later and remained engaged until time 05:39:55. By that time MCAS had completed 3 discrete pitch-down trim inputs before the AP actually disengaged.
George the autopilot wants to go up whilst HAL the MCAS wants to go down.
Eeek.