It has recently been argued that the distribution of waiting times between
successive solar flares is incompatible with the prediction of lattice mode
ls, which interpret flares as avalanches of magnetic reconnection events wi
thin a stressed magnetic structure driven to a state of self-organized crit
icality by stochastic motions of the photospheric magnetic footpoints. Insp
ired by a suggestion recently made by Wheatland, we construct modified latt
ice models driven by a nonstationary random process. The resulting models h
ave frequency distributions of waiting times that include a power-law tail
at long waiting times, in agreement with observations. One model, based on
a random walk modulation of an otherwise stationary driver, yields an expon
ent for the power-law tail equal to 2.51 +/- 0.16, in reasonable agreement
with observational inferences. This power-law tail survives in the presence
of noise and a detection threshold. These findings lend further support to
the avalanche model for solar flares.