Sure. We always laughed at the 3k orographics and knew it was overdoing it in the mtns…it was more of a signal rather than taking the QPF verbatim. The 12k always seemed more zonked in coastals to me even near go-time. The running joke here in the winter is we’re always waiting for that run the day before a SNE biggie that jackpots here to northern VT. But the 3k was really good with CAD and (imo) ptype in those overrunning situations. Convection is hit or miss, but I never saw it as any worse than the other hires models. I just feel like it had its purpose and times where it was useful.
We still seem reluctant to dive into full 4dvar like ecmwf. I haven’t talked to Alicia B. in years so I should pick her brain more, but maybe you can explain to me while we’re still messing with 3.5dvar or other hybrids. I remember a study many years ago that showed the gfs error scores were similar to the ec when using their data assimilation. So it seems like there’s value there.