Basically, if you want multiple sounds playing at the same time (e.g. music + desktop notifications) and you don't have hardware mixer (many common sound chips don't and rely on Windows drivers/sound system to provide the functionality), you are out of luck (no, the ancient dmix plugin is not a solution!). Bluetooth audio (headsets) doesn't really work without PA.
Configuring apps to use different devices (music should go to speakers and video conferencing to a headset ...) is a pain without PA - most applications don't let you select the input/output devices.
And many other things. ALSA is good, but without PA the sound support on a modern Linux desktop would be stuck right in the late 90s. Is it necessary? Not strictly, but it is one heck of a convenience that most don't even realize they have.
Here is a good post from 2008 explaining many of the issues PulseAudio solves and addressing some of the old FUD:
http://0pointer.de/blog/projects/jeffrey-stedfast.html
Unfortunately, there are tons of people with strong opinions about both PA and systemd but very little actual knowledge about what these components do and what issues they address in a modern Linux/Unix system. But that doesn't prevent them from spreading BS FUD and conspiracies about this or that group trying to dominate the market or take over the competing Linux distros.
Given that that's exactly what it's for, why isn't it?
Why write for ALSA when the distro ships with PA, and so all of its configuration tools are PA-focused?