Hypothesis: This Epic, together with Solr/EZ, Profiling+Performance efforts, Improved SQL engine, and potentially also Symfony 3.0 performance efforts, are the efforts needed to be able to reach really good levels of performance for uncached requests, assuming Cluster implementation can scale up this allows for many new use cases with eZ Platform that has previously been difficult/impossible to handle.
HHVM, the PHP VM made by Facebook provides far higher performance then PHP 5.x, especially for heavy applications like eZ Platform.
PHP 7.0 coming end of 2015 will reduce some of the gap here, so supporting HHVM should rather be about providing the option for customers, and in given HHVM will continue keeping their performance lead even once PHP 7.0 is out.
Possible followup to this Epic is to selectively take advantage of Hack, the improved PHP language available on HHVM, to improve performance further by getting rid of some of our Magic getters and setters that we need for readonly properties for instance. For PHP we can consider making PHP extension for same purpose until such feature is natively supported by the language.