Just to illustrate why the annual, premature trumpetings of ice recovery are frustrating - a sharp decline in May has put the Arctic ice extent at record a monthly low;
so the picture now looks very different. Whether that sharp decline continues is, of course, anyone's guess; we don't know which way the wind will blow - but it will be warm this year, the ice isn't likely to grow back before winter and the higher ratio of fragile single year ice resulting from previous severe melts suggests that it's likely that this year's summer minimum might be very low again - eating further into the more stable multi-year ice. These very low years do have very real long term implications for arctic ice extent and the Earth's albido.
---
Hi Digit.
It's really not an easy question to answer - because (apart from me not being an expert - At best I'm an OU student, at worst an armchair expert) the Global Mean Surface Temp is a measure of energy in the climate system, not a measure of what the weather is like. What is being gauged is the energy available to drive weather systems and patterns - things like arctic oscillations, gulf streams, el-ninos and la-ninas. How those things work together to produce the weather (or even the local climate) we experience given more or less energy is beyond my powers of reckoning.
You also have to take into account that the Thames has changed physically - it was wider and slower in the C19th than today. And I'm also not sure that frost fairs were 'normal' rather than exceptional; Wikipedia thinks that there are only records of the Thames freezing over at London on
24 occasions in the last 400 years - and they seem like pretty newsworthy events when it happens. It might be that those really harsh winters that witnessed frost fairs were exceptionally cold - not part of a natural trend toward plummeting temperatures.
Industrial emissions of CO2 grow exponentially, or there abouts. Human industrial activity grows year on year and atmospheric CO2 accumulates - so each year it builds on the previous. It's the nature of steady state growth (as opposed to linear growth) to be pretty insignificant in its early stages only to kick in later (think doubling grains of rice for each square of a chess-board) - so I'm not sure there's much reason to think that the world would be doing much of anything really noticeably different to what it actually did before the mid C20th in terms of reacting to CO2 emissions. The increase in GMST in the first half of the C20th is readily attributable to natural influences.