@adiz @mischievoustomato @moshibar wavpack has 'hybrid' mode where you get two files: one that uses lossy compression, and a residuals file, which if both are present you get a lossless decode otherwise you get the lossy one.

32-bit float is something thats mostly just used internally by DAWs, or by high end recorders nowadays because its precise enough you don't need a gain control anymore. you just fix the volume in post.