LM2588: Chip randomly starts up in stuck mode

I have a simple flyback transformer based boost supply built around an LM2588, and noticed it would occasionally start up in a mode where the circuit would draw excessive current.  Originally, it would recover after a short while and operate normally, but while debugging the problem, I tried adding a separate power supply for the LM2588.  Sometimes it will start up and work properly, but other times, it ends up in one of two modes where the SW pin stays low most of the time (drawing high current, saturating the transformer, and producing low output voltage).  I have a 22k resistor to the frequency adjust pin to set the operating frequency to 200kHz.  When it's working, it works nicely at this frequency.  When it gets stuck, one mode is at 200kHz, but the "compensation" pin is at about 2.6V, the other stuck mode is at about 53kHz, with the "compensation" pin at about 0.4V.  I've tried several of the remedies in the data sheet, including adding a Schottky diode to clamp negative transients on the switch pin, changing the compensation network, and adding an RC low-pass filter to the Vin pin.  None of these avoid the problem, it still powers up in one of the stuck modes occasionally, and now that it has a separate clean Vin supply, it doesn't recover (I think it recovered initially because the power supply would current limit to 1.5A, which would drag Vin down from 12V to about 4V, whereupon it would shift to a different mode and begin working properly after 500ms - 10s or so).