Why one number is not enough
A frame with a single 4,000-nit specular glint and a frame that is entirely a 4,000-nit white field have the same MaxCLL and are completely different problems for a display. The first is trivial; the second draws far more power and generates far more heat than most panels can sustain.
MaxFALL is what separates them. It reports the highest frame average, so a display can apply automatic brightness limiting where it is genuinely needed instead of clamping every bright highlight defensively.
Typical values
MaxFALL is almost always far below MaxCLL, because real frames are mostly midtones. A clip with MaxCLL of 1,000 might carry MaxFALL of 150 to 400. A MaxFALL close to MaxCLL means a sustained bright image, a snowfield or a white studio backdrop, and that is where panel limiting becomes visible as the picture dimming a second after a cut.
Reading it
The same ffprobe side-data query that returns MaxCLL returns MaxFALL as max_average. See MaxCLL for the command.
Questions people ask
Why is MaxFALL always lower than MaxCLL?
Because real frames are mostly midtones. MaxCLL reports the single brightest pixel and MaxFALL averages a whole frame, so unless the frame is a uniform bright field the average sits far below the peak. A clip with MaxCLL 1,000 might report MaxFALL between 150 and 400.
What if MaxFALL is close to MaxCLL?
That means a sustained bright image: a snowfield, a white studio backdrop, a full-frame flash. Panels limit brightness under those conditions for thermal reasons, which is what you are seeing when a picture visibly dims a second after a cut to white.