It's a gamble. On one hand, you may only spend a few hundred bucks on stuff like 12v batteries or charging ports. But on the other, your DC/DC converter or touchscreen might go and you're out $4k right there--the cost of the ESA. If both go? You'll be kicking yourself. Then there are reports of the 12v batteries failing pretty much annually, so that's another $400 per year.
If Model 3 is truly simpler and better constructed than Model S/X, we shouldn't have to worry about the niggling issues like door handles, 12v batteries, squeaks, leaks, etc. and you may come out on top even with a major out-of-warranty repair like the touchscreen. But it'll be close.
Ultimately, I think the ESA is worth it mainly because I'm not a huge risk-taker and am quite intimate with Lady Luck's evil twin. A brand new first gen car? There will be issues down the road that aren't caught right away because they don't present themselves until years later. Let's be real here, modern car companies who have been making the same damn car for 100 years still can't get it right. Do you honestly believe that a 10-year-old tech startup whose focus is batteries and technology (not cars) will perfectly execute the first car of its platform without a hitch?
You may be lucky and only have to spend the couple hundred here and there for the additional 4 years, but when a single repair of a 5+ year old part that has zero reliability data costs as much as four years of ESA? I'm not willing to take that chance. Good to luck to you if you do.