Notes. The columns contain the following values: t − T₀ shows the temporal difference between the optical maximum T₀ and the event time in units of hours; T₁ and T₂ are the start and end times of the feature in units of MJD+57 000; ΔT is the duration in units of s; a₀ is the fitted amplitude of the nova model; χ²∕ν is the goodness-of-fit value, weighted by the number of degrees of freedom (reduced χ²), indicating whether a nova model is appropriate to express the light curve^b; is the significance of the feature over the background in units of σ; and HR₅₀₀ is the hardness ratio difference, comparing the times before, during, and after the event in units of σ. The last column provides plausible source identifications, based on the other columns, coincidences with GOES X-ray data, or the INTEGRAL response. The horizonal dashed lines indicate times for which the gamma-ray flash is “most probably” expected to occur (Gomez-Gomar et al. 1998), i.e. between days 2 and 10 before the visual maximum. ^(a) Two events which are coincident within a few minutes are captured by the search method as only one. This happens for example when a strong GRB precedes a solar particle event. Here, upper limits are given for fluxes and significances. ^(b) Because the nova light curve model is not parametrised and thus is very stiff, the resulting χ² values are subject to extreme values and variations. We take this into account by also considering apparently bad χ² values, up to 25σ deviations from the optimum. This declares reduced χ² values below ≈ 4 as good, given the number of degrees of freedom of around 160. ^(c)Intriguing events discussed in the main text although formally excluded by specific diagnostics.