There's no simple answer to this; consider that, if you enable BGP on a Hub MX, to advertise your branch subnets into the Data Centre, you will also be learning routes from the eBGP peer in the DC, which will then be advertised to all the Spoke MXs of that Hub. As the Spoke MXs tend to be 'smaller' units than Hub MXs (e.g. MX67, compared to MX250) - and because DCs often learn lots of routes, it is generally easier to overload the Spokes than the Hubs.
Look to limit the routes advertised from the upstream eBGP peer and summarise where possible.
Look for <150 routes learned via eBGP, to protect the spokes
Unless you have an unusually large number of subnets, per branch, if you follow the scaling guidelines in the MX Sizing Guide - particularly for tunnels & VPN throughput in particular - I think you're unlikely to hit a scaling limitation on the Meraki Hub end. Also look into the impact of the routes advertised to the upstream eBGP peer in the DC.
