Fred nailed this one right on!
If one checks a glossary in a reference book, one will find -
proof load
A predetermined load, generally some multiple of the service load, to which a specimen or structure is submitted before acceptance for use.
However,
proof stress
(1) A specified
stress to be applied to a member or structure to indicate its ability to withstand service loads. (2) The stress that will cause a specified small permanent set in a material.
And it is the second definition which applies here.
IIRC, 0.1% strain offset has also been used in addition to the 0.2% offset, in the US. According to ASM's "Atlas of Stress-Strain Curves", in the UK (and perhaps outside of North America), offset yield strength is referred to as proof stress, and 0.1% or 0.5% is used.
The point of choosing 0.2% is that it is easier that 0.1% and not as much as 0.5%. Using the offset "avoids the practical difficulties of measuring the elastic or proportional limit" of a material. This is not so much a problem these days with digitally controlled tensile test systems, but it was significant 25 or more years ago before high quality digital electronics were available.
The idea was to get as close as reasonably possible to the elastic limit.