When HTML was first designed, the NextStep development environment included a rich text editing component, and SGML was popular in the hypertext community; in particular, the IBM GML starter tag set was in use at CERN. Tim Berners-Lee focussed on adding hypertext linking; he didn't spend a lot of energy on elements such as p, h1, and li, but it was clear that the list he started with would naturally grow over time.
Therefore the original specification and implementation of HTML had a rule to IgnoreUnknownTags.
This worked well for elements such as em and strong, where the impact of ignoring them was, in many cases, an acceptable reduction of fidelity.
For elements such as form and table, the reduction in fidelity is less often acceptable, but even in those cases, while the reader might miss information in the document, the miscommunication isn't as costly as switching red and green on a traffic signal light.
But the IgnoreUnknownTags rule has serious drawbacks as well: [@@tell story about bad side of tag wars and such..]