Events of extreme intensity in turbulent flows from atmospheric to industrial scales have a strong social and economic impact, and hence there is a need to develop models and indicators which enable their early prediction. Part of the difficulty here stems from the intrinsic sensitivity to initial conditions of turbulent flows. Despite recent progress in understanding and predicting extreme events, the question of how far in advance they can be ideally predicted (without model error and subject only to uncertainty in the initial conditions) remains open. Here we study the predictability limit of extreme dissipation bursts in the two-dimensional Kolmogorov flow by applying information- theoretic measures to massive statistical ensembles with more than 107 direct numerical simulations. We find that extreme events with similar intensity and structure can exhibit disparate predictability due to different causal origins. Specifically, we show that highly predictable extreme events evolve from distinct large-scale circulation patterns. We thus suggest that understanding all the possible routes to the formation of extreme events is necessary to assess their predictability. This dataset includes the GPU code used to perform the simulations of two-dimensional Kolmogorov flow (code.zip) and the database generated for the analysis presented in the paper (database.zip). The database contains the time series of the spatially-averaged dissipation for 8192 independent simulations (termed base trajectories). For each base trajectory, an ensemble of 8192 simulations was assembled in which the initial condition of the base trajectory was perturbed with a small random disturbance and then evolved. Thus, the database contains 8192 x 8192 temporal evolutions of the dissipation which allow testing the predictability of extreme dissipation events. A script is provided that exemplarily reads a base trajectory and plots it together with 10 perturbed evolutions of the ensemble.In addition, for 1040 base trajectories featuring an extreme event, the time series of the contributions of the horizontal and vertical large-scale modes to the dissipation are provided (extremes.zip), enabling to study the effect of the large-scale structure of the flow on predictability.