Currently it is not supported to have different algorithm inputs for continuous children with the same PID. However this can be an issue for long-lived particles, when one candidate is reconstructed from long tracks and the other one is reconstructed from downstream tracks (see for example Moore!1249 (merged))
Here I try to fix the issue, enabling this possibility for a limited number of predefined particles (thanks to @mstahl for the suggestion). @mvesteri @poluekt @nnolte @rmatev can this be a viable solution?
cc @lpica