This is quite interesting. Keeping in mind that social grade is a bad proxy for class, something has happened to voter alignment around redistributive policies.

(1) hypothesis one: realignment + partisanship. Voters instinctively identify the policy as a pro-Labour, anti-government proposition and react to it according to their partisan priors, which have become a lot less class-based than before.

(2) hypothesis two: social values. Via education, social liberalism is now prevalent among the rich, and it informs some aspects of economic attitudes: i.e. redistribution directed towards a 'deserving outgroups'. Plus it comes with higher awareness of social desirability.

(3) hypothesis three: the specific politics of free school meal. Because the measure is not immediately linked to taxation and macroeconomically trivial, there's a lesser 'selfish' dimension to it. It is informed by values of non-economic nature, rather than economic interest.

(4) hypothesis four: age distribution of occupations. Because younger adults are more likely to be parents than the elderly, and the former entered a job market with more 'ABC1' jobs than 'C2DE' jobs, there are actually more financially insecure parents as a share of ABC1 voters.

(5) hypothesis five: this is a subgroup analysis of one poll, which uses a questionable social grade measure, so it's virtually meaningless.

That said, I think it's important to look at how the distribution of economic attitudes has changed in society from, say, 30 years ago. It's not just social attitude shifts or across-the-board changes in economic policy mood: there's something else moving at a glacial pace here.