Why do you want to force new lines in the middle of paragraphs ? depending on how much width there is, that may be sub optimal

I think that:

quant_table_set_count indicates the number of Quantization Table Sets. Inferred to be 1 if not present. MUST NOT be 0.

is less readable as:

quant_table_set_count indicates the number of Quantization Table Sets.
Inferred to be 1 if not present.
MUST NOT be 0.

As each line is a different part of the definition. Having a different paragraph for each line would be too much.

another problematic example is:
https://github.com/FFmpeg/FFV1/blob/master/ffv1.md#initial_state_delta

initial_state_delta [ i ][ j ][ k ] indicates the initial Range coder state, it is encoded using k as context index and pred = j ? initial_states[ i ][j - 1][ k ] : 128 initial_state[ i ][ j ][ k ] = ( pred + initial_state_delta[ i ][ j ][ k ] ) & 255

It is very difficult to read (for me: it is impossible to understand, I just see a single formula with two = and I don't understand the link between each part), with pred definition and initial_state[ i ][ j ][ k ] definition on the same line, without any separator, but the source is:

initial_state_delta [ i ][ j ][ k ] indicates the initial Range coder state, it is encoded using k as context index and
pred = j ? initial_states[ i ][j - 1][ k ] : 128
initial_state[ i ][ j ][ k ] = ( pred + initial_state_delta[ i ][ j ][ k ] ) & 255

From my point of view the current version (all on 1 line, same separator between a part of the formula and another formula, i.e. space) is not understandable as a definition of of pred then a definition of initial_state.

This is definitely very subjective, we can abandon this change if nobody else thinks it is useful. The opposite idea would be to remove the carriage returns in the Markdown source as the Markdown version makes thinking that there is a carriage return where there is no carriage return.

Maybe at least keeping modification for formulas (pred / initial_state formulas issue)?

From my point of view the current version (all on 1 line, same separator between a part of the formula and another formula, i.e. space) is not understandable as a definition of of pred then a definition of initial_state.

The formular in your example a bit slow to parse in both cases. There are 4 things the text conveys to the reader, 1. what it is, 2. what is stored (delta mod 8 from some predictor), 3. the predictor, 4. the context used for the coder. I agree the newline makes it better in this case but it still intermingles things and isnt very clear

Status of this? It appears that @JeromeMartinez's addition of spacing for line breaks makes it more readable. @michaelni points out that it could be more readable, but that could occur in a subsequent commit and doesn't have to block this.

Sorry, I inadvertently deleted the branch.
Pinging about this PR, if not acceptable a is I would like to have hints about what would be acceptable.
My opinion is similar to the one from Dave (changing even more in order to be more readable would be part of another PR, this one is for carriage returns only).

this doesnt apply cleanly anymore

Branch rebased.