If you P-lock the rate at the same time as the random LFO (also set to rate), this sets the centre point for the LFO. The LFO does override this but still uses the p-locked value to determine its centre.
For example, P-lock rate 48, random lfo to rate with depth 32. This means the lfo will never drop below 32 or exceed 64. This gives the result I was originally looking for, a randomly mashed up beat that never misses playback trigs (BUT it only ever drops to half speed)
By taking the lfo lower than half speed (below rate 32) you introduce a probability of the sample not firing. For example the settings in my original post give a randomly mashed beat but with 50% chance of not playing back at each trig. Or another example, P-lock rate to 40, random lfo to rate with depth 48, this makes the lfo work between rate 16 and 64 with a 33% chance of the sample not firing at each trig.
Im still not entirely sure why the sample doesn't trigger when the rate is below 32 but thats how it works when manipulating the recorder buffers live. I hope this info will be useful to someone! 