Comment (by Zach Etienne):

The “Duez et al” trick is quite powerful – I implemented this trick in Baikal a few months ago and found H constraint violations drop by about an order of magnitude compared to when it’s not enabled, for binary black hole simulations (e.g., a ET BBH gallery example - like simulation).

The trick dates back to Yoneda & Shinkai https://arxiv.org/abs/gr-qc/0204002 , who did an analytical (non-numerical) investigation

… and Duez et al I think were the first to report its use in NR simulations: https://arxiv.org/abs/gr-qc/0209102

In addition Raithel & Paschalidis recently found that this trick improves H constraint convergence behavior in the context of binary neutron star simulations: https://arxiv.org/abs/2204.00698

Future versions of Baikal/BaikalVacuum will have this trick enabled by default – I’m now calibrating the magnitude and scaling of the coefficient.

--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/963/improve-mclachlan-accuracy