I tried to simulate, using the Mercury integrator by John E. Chambers, their past relative distance and velocity.
The data are still very uncertain and so this simulation is by no means conclusive: however, it seems that in these days these two asteroids are at the end of a much longer cycle where they happen to come relatively near.
From file info.out:
Algorithm: Bulirsch-Stoer (conservative systems)
Integration start epoch: 2458000.5000000 days
Integration stop epoch: -100000000.0000000
Output interval: 100.000
Output precision: medium
Initial timestep: 0.100 days
Accuracy parameter: 1.0000E-12
Note: the apparent zero points both in relative distance and velocity are NOT really zero!