“…Most importantly, we cannot properly build and evaluate meta-solvers that schedule more than one individual solver in the solving time window [0, τ ), because we do not know the best solution found by a solver at a time point t < τ . This is unfortunate since scheduling different solvers is very common for a number of effective meta-solvers (Hula, Mojzísek, & Janota, 2021;Gonard, Schoenauer, & Sebag, 2017;H. Hoos, Kaminski, Lindauer, & Schaub, 2015;Amadini, Gabbrielli, & Mauro, 2014;Malitsky, Sabharwal, Samulowitz, & Sellmann, 2013;Xu, Hutter, Hoos, & Leyton-Brown, 2008).…”