Occasional test failures of hlt2_reco_calo_selectivematching

I think this test randomly fails every now and then with a single diff:

    (SelectiveBremMatchAlg ref) average chi2 | 64109 | 16728.91 | 0.26094 | 0.40398 | 7.3711e-11 | 17.781
    (SelectiveBremMatchAlg new) average chi2 | 64108 | 16728.96 | 0.26095 | 0.40398 | 7.3711e-11 | 17.781

Take for example the lhcb-master-ref slot, test passed in 1311 and 1313 but failed in 1312.

Could it be that we have a race condition or a use after free in that algorithm?

fyi: @cmarinbe @graven @mveghel