> The stupid idea was basing the meter on the distance between the North Pole and the Equator
Why is that stupid?
The point of SI (and its predecessors like MKS and CGS dating from the early 19th century) is readily achievable, coherent and replicable standards for not just length, but also mass, time, temperature, electric current, amount (of a uniform subsatance), luminous intensity and several derived units like volume (of an arbitrarily shaped container), pressure, electric charge, force, and so forth.
Some of the fundamental units were harder to insert into a coherent system in an easily replicable and achievable way.
You've chosen to look at two such units -- length and temperature.
The metre has an interesting history whose beginnings suffered from difficulties in achieving independent reproducibility of the standard metre. Starting in the 17th century, various approaches were explored, with two leading candidates surviving into the 19th century, both relating to the geometry of the Earth in principle measurable everywhere with suitable equipment. One candidate required a detailed survey of a meridian and an almanac of angular measurements one could make against objects in the sky or objects receding over the horizon or alternatively with a map angular measurements of objects of known height disappearing over the horizon. Another candidate required an excellent portable frequency standard and an almanac relating that to time-of-day in a location-dependent fashion.
The first is the version that survived until the definition of the metre was tied to the properties of atoms and the universality of the speed of light, mostly because it was more reproducible. This was the "meridional" version; one could readily produce a high precision metre prototype with good equipment and a stable platform on a large calm body of water extending to the horizon along a meridian (this means one could do so essentially on the shore of a large lake). The definition is 1/10000 of 1/4 of a great circle through both poles of a geoid that averaged out slight differences in oblateness of the Earth's surface. Apart from error terms relating to the non-uniformities in the Earth's true surface, realizations using trigonometry against objects in the sky suffered uncertainties because of the several sources of variation in the rotational speed of the planet. With the advent of GPS and decent approaches to defining a working geoid, a "meridional" approach is still viable, although I would be surprised if anyone seriously proposed dong so as a replacement for the present definition based on the speed of light.[1]
The other leading candidate was the "pendulum" method. When one constructs any pendulum whose half-period is one second anywhere on the surface of the planet, the arm of the pendulum will have a length very close to one metre. One critical problem is that local mass concentrations, altitude, and latitude all influence the length, and already in the 18th century it was clear that the length of the pendulum could vary by several millimetres within a radius of even a few hundred kilometres in some places, and there was no a priori way to determine all of the local contributions in order to achieve the same accuracy possible with the meridional method. Worse, precisely calibrating a seconds pendulum was a difficult technical challenge even in laboratories, even though the underlying mathematical formula was fairly simple. The problem is that in ideal situations the dominant driving term is "g_0", the standard acceleration of terrestrial free fall (as it is now known). Unfortunately the actual acceleration of objects near the surface in free fall in vacuum varies significantly across the whole of the planet, and can even vary over relatively short timescales at one location, and there is no a priori way to determine the expectation value with great reliability. Indeed we've had relatively poor data with which to build a global almanac until the 21st century with satellite observatories like GRACE and GOCE[2], and even now relying on a seconds pendulum for defining a metre is an unattractive proposition.
The metric unit of temperature is the kelvin, not the degree Celsius. The Celsius scale is based on kelvins, but with a 0 point (that of the triple point of a particular standard of purified water at a particular pressure) that is fairly straightforwardly achieved with decent precision even in a typical school science classroom setting. The kelvin is defined in terms of of an exact fraction that triple point, and it "only" suffers some difficulties in the exact definition and realization of 0 K.
The kelvin is due to be redefined in the next few years taking a fixed value for the Boltzmann constant k_B which can be expressed in terms of J K^-1 where J is Joules and K is kelvins, while the modern Joule is already defined in terms of the Planck constant h, the speed of light in vacuum, and the second. This redefinition is aimed principally at coherency as mentioned at the top, as we replace features common near the surface of the Earth everywhere people live with physical constants expected to be the same everywhere in the observable universe. A strong parallel goal is reproducibility and realizability of the units; the kelvin was already easily reproduced with high precision, and few metrologists are wholly comfortable with a definition that could be much harder to "show" in a lab or factory.
> "more consistent"
The Kelvin scale has always been well-integrated with the other units of SI and its predecessors, particularly since the beginning of the late 19th century programme of defining units in terms of universal physical constants.
The equivalent in U.S. Customary Units is the Rankine scale, which has the same 0 point as the Kelvin scale, but using Fahrenheit-sized degrees (which in the U.S. are anyway defined by NIST as 5/9 K, and NIST prefers "rankine" over "degree Rankine"). There is no (formal) "Imperial" rankine; AFAICT the whole of the former British Empire uses kelvins either in the Kelvin scale or (when discussing weather or cooking, for instance) the Celsius scale, although proximity to the USA and aborted-by-1980s-politics conversion to metric leads to Canada using a mix of units -- e.g. Celsius in weather reporting and Fahrenheit in household cooking.
However, given the 5/9 constant conversion factor, using rankines vs kelvins is a matter of choice. The placement of a zero point in a scale with such degrees is almost essentially arbitrary (although there are obvious attractions for some realizable ground state as the zero point. While "absolute zero" might be reached asymptotically with close approximations of an ideal gas it is not an obviously perfect choice on theoretical grounds), and is a matter of suitability. Thus there is no clear "better" between the two everyday temperature scales; each has advantages and drawbacks. Equally importantly, neither offers scope for improvement of definition of degree or zero than the other.
Essentially all scientific applications use the Kelvin scale; most of the world is comfortable using Celsius in non-scientific applications. There is almost no use of the Rankine scale (even in the USA), and most people in the USA are comfortable using Fahrenheit in non-scientific applications. Some cultures use a mix of Celsius and Fahrenheit in everyday situations. And of course, many cultures have never used the Fahrenheit scale. None of these cultures or the economies they participate in seem to be on the verge of collapse because they have made a "wrong" choice of temperature scale, however it is notable at how quickly the everyday use of Fahrenheit in most of the former British Empire collapsed.
- --
[1] advances in (long-baseline) interferometry in the late 19th century could have been directed at improving the definition of the "meridional" metre, but it would have been odd not to take advantage of short-baseline interferometry that is at the heart of the "wavelength" definition proposed by Michelson (of the famous Michelson and Morley experiment). Unfortunately a consistently reproducible monochromatic emitter was unachieved until the middle of the 20th century, and even with the advent of solid-state lasers, the "wavelength" interferometry approach has more sources of uncertainty than the present light-second definition.
[2] http://news.nationalgeographic.com/news/2011/04/110406-new-m...