Opera wasn't a browser from scratch, though: Presto was older than KHTML (and WebKit, etc.).
And really, starting from scratch ten years ago you'd just run into Presto's big failing: site compatibility! Websites rely on all kinds of asinine edge-case behaviour, and if you don't match the majority behaviour, users will leave your browser for your competitors (and site compatibility was, along with crash bugs, always one of the top two reasons for users to switch away from Opera).
In many ways, the vastly larger HTML and CSS specs are a massive boon for minority browsers: when I started at Opera, a large proportion of the Presto team were QA staff who in reality spent almost all their time reducing site compatibility bugs and reverse-engineering other browsers. HTML5 and CSS2.1 made to a large degree that work go away: there was enough movement to converge behaviour (including from the larger browsers) on documented behaviour that reverse-engineering other browsers ceased being something consuming large amounts of resources on all browser teams.
What killed Presto is a variety of things, and the growth of the platform was only a small part of that.
And as mentioned in other sibling comments, all major browsers have rewritten major components at various occasions.
Major browser dev teams rewriting components over year-long periods of carefully planned integration points (with Mozilla even introducing a new programming language along the way) is hardly telling anything about the viability of developing a browser from scratch. Given the powers-that-be in so-called "web standards", by the time you've got anything to show odds are it'll be obsolete.
What would be helpful is if the whole web stack could be organized into profiles (eg. things working without JavaScript, without CSS "4" features, etc.), but WHATWG dismissed the idea of HTML/CSS versions or device profiles proposed by MS (hell they even can't be bothered to version their "living standard" specs). And W3C could start to give formal semantics for CSS (which is kindof "implementing" and verifying the spec) rather than prose about dozens of layout models in an untyped ad-hoc syntax. That is the role of standard bodies, not to lead us into a way-of-no-return for the benefit of very, very few people.
Opera wasn't a browser from scratch, though: Presto was older than KHTML (and WebKit, etc.).
And really, starting from scratch ten years ago you'd just run into Presto's big failing: site compatibility! Websites rely on all kinds of asinine edge-case behaviour, and if you don't match the majority behaviour, users will leave your browser for your competitors (and site compatibility was, along with crash bugs, always one of the top two reasons for users to switch away from Opera).
In many ways, the vastly larger HTML and CSS specs are a massive boon for minority browsers: when I started at Opera, a large proportion of the Presto team were QA staff who in reality spent almost all their time reducing site compatibility bugs and reverse-engineering other browsers. HTML5 and CSS2.1 made to a large degree that work go away: there was enough movement to converge behaviour (including from the larger browsers) on documented behaviour that reverse-engineering other browsers ceased being something consuming large amounts of resources on all browser teams.
What killed Presto is a variety of things, and the growth of the platform was only a small part of that.
And as mentioned in other sibling comments, all major browsers have rewritten major components at various occasions.