I haven’t read this imperial college paper yet, but yesterdays preprint from SA has it at 80% less likely to lead to hospitalization, with some uncertainty as to whether that’s due to intrinsic lower virulence vs underestimated prior immunity

https://doi.org/10.1101/2021.12.21.21268116

Edit: read through a good portion of it, the methodology seems pretty reasonable, including definitions for infection/reinfection, inclusion criteria, the covariates included in the model. I’d say the biggest uncertainty factors are (1) it’s still early and there aren’t that many hospitalizations yet (the study period ended like a week ago) and (2) for their correction accounting for undercounted prior infections, they assume undercounting of 1/3 based on some other papers. Maybe that’s reasonable, I’m not that familiar with testing in the UK. But they report both corrected and incorrect results

I'd be pretty concerned at the really small date range for this study, but otherwise it looks reasonable. The corrections of hazard ratios is the part that might, if you needed to, be hiding some trickery.