Thanks for the feedback! We'll certainly work on the points you mentioned, although validation is no goal in itself for us. We just hired our first frontend developer actually, so far we are still recycling the original (very well validating) templates an external agency made for us, but of course over time they've "broken" in terms of validation.
Optimizing is something we'll do for sure in our redesign since we'll redo all the HTML anyway. My guess is that it won't be a hefty difference as you suggest, since our pages already load in 1,6s on average for NL users (via Pingdom).
We can go a bit further in caching with Nginx too btw.
On your last point: to a certain extent it's true that the SEO or other marketing guys can also make tests. But in practice, for the most interesing tests you really need a frontender, the basic stuff you can do with VWO and some CSS hackery doesn't bring us much anymore.
Be sure that we try to eat our plate, but we won't eat it for the sake of emptying it. Desert might be more efficient, and there's always the next meal (day) with new opportunities that need chasing. You're 100% right in theory, but not in practice :)