What do you think of thunks?

It appears one can't edit the original post so I am replying to myself in order to update the workbook.

The key difference is that, although I have demonstrated the results of 9 distinct, but related, formulas, there are only two formula cells.

By bringing the calculation together within a single module, it became obvious that there was a lot of repetition, involving evaluating thunks and applying a in-built function to the result, functions that could be tidied up by writing a further Lambda function

CALCλ
= LAMBDA(values, n,
    LET(
        // Call ROLLINGRANGEλ to extract a rolling range of n cells from the range 'values'
        arrayϑ, ROLLINGRANGEλ(values, n),
        // Expand the thunks to give a 2D range for demonstrtion purposes
        array, TRANSPOSE(EVALTHUNKARRλ(TOROW(arrayϑ))),
        // demonstrate that the thunk array arrayϑ contains ranges and not simply arrays
        isref?, MAPϑ(arrayϑ, ISREF),
        // demonstrate that the ranges are of differing dimensions
        rows, MAPϑ(arrayϑ, ROWS),
        // calculate the average value over each range to give a rolling average
        avrg, MAPϑ(arrayϑ, AVERAGE),
        // calculate the maximum value over each range to give a rolling maximum
        rmax, MAPϑ(arrayϑ, MAX),
        // compare each value to its rolling maximum
        isMax?, values = rmax,
        // Stack reults for presentation
        HSTACK(date, array, isref?, rows, avrg, rmax, isMax?)
    )
);

which now use the function in order to combining the thunk evaluation with a sequence of aggregations

MAPϑ
= LAMBDA(arrayϑ, FNλ,
    MAP(arrayϑ, LAMBDA(ϑ, FNλ(ϑ( ))))
);

I realise that this style of programming is not going mainstream anytime soon, but the fact that it is possible and appears to work well should be of interest.