1. There is a lot of “math” in digital signal processing and certain things (like filtering) require summation. The programming is easier in floating-point and big numbers (from summing) are not a problem. Virtually all audio editors/DAWs work in floating point.
2. The “maximum count” in integer audio is defined as 0dB and it can’t go higher.* If you have a typical 0dB normalized/maximized file and you boost the bass (or anything else that boosts the volume) you’ll go over 0dB. Integer audio would be hard-clipped at 0dB and permanently distorted. With floating-point there is essentially no upper (or lower) limit so you can go over 0dB temporarily and then reduce the volume to a “safe level” before exporting.
Note that Audacity “shows red” for potential clipping. It doesn’t really know if the waveform is actually clipped.
Also, digital-to-analog converters (playback) and analog-to-digital converters (recording) are integer-based and they will clip if you play a floating-point file that goes over 0dB at “full digital volume” so your finished files shouldn’t go over 0dB even if you use floating-point WAV.
\
Everything is automatically scaled during playback (and recording) so although a 24-bit file has bigger numbers than an 8-bit file, they play at the same volume.
FLAC files are nearly always either 16-bit or 24-bit.
8-bit and 12-bit FLAC files are also possible, but hardly ever used.
Audacity supports importing any valid FLAC file (8 to 24 bits per sample).
Audacity supports exporting 16-bit and 24-bit FLAC.
On import, Audacity decodes FLAC files to 32-bit float PCM. This conversion is lossless.
If you mean, what’s the advantage of the additional 8 bits (32-bit float vs 24-bit integer), the main benefit is that 32-bit float can go over 0 dB. This can be a major advantage during production because clipping at 0 dB (which all “integer formats” do) can cause irreparable damage.
There may also be a very slight performance advantage in using 32-bit float during production as 32-bit float is a “native” number format in computers.
When recording at 16-bit, you need to be more careful than when recording at 24-bit. In the latter case you can safely allow a lot more headroom, but with 16-bit the headroom should be no more than 6 dB for optimal quality.
With 24-bit recording you can probably allow a lot more than 6 dB headroom, if you want to, with no audible loss of sound quality, though it does depend on the quality of the equipment.
Somewhat oversimplifying, but as a general idea:
The theoretical maximum dynamic range for 24-bit audio is 144 dB. If, for example, you have a microphone / pre-amp combination that have a dynamic range of 100 dB, then theoretically you could peak at -44 dB and still retain the full dynamic range of the analog hardware.