It was pointed out that chopping a cube into slices over a dimension and re-merging (which was needed modify each part individually) had the effect of reversing a dimension coordinate with descending values.

And such a reversal breaks the 'is_contiguous' property :

E.g.

>>> co_upwards = DimCoord(points=[20, 30, 40], bounds=[[15, 25], [25, 35], [35, 45]])
>>> co_ upwards.is_contiguous()
True
>>>
>>> co_downwards = co_upwards[::-1]
>>> co_downwards.is_contiguous()
False

Because what ‘contiguous’ means is just "bounds[i, 1] == bounds[i+1, 0], for all i"

According to : cf 1.7, section 7.1,
There under "Bounds for 1-D coordinate variables" we have :

If adjacent intervals are contiguous... For example, if the intervals that contain grid points lat(i) and lat(i+1) are contiguous, then latbnd(i+1,0) = latbnd(i,1)

So, our implementation of 'is_contiguous' does seems to be in line with the CF conventions,
but our reversed (dimension) coordinate effectively has invalid bounds,