Because year after year, the new version of Perl was supposedly on the way. After they pretty much killed the language by having a new version that wasn't ready and nobody wanting to invest in the old version, they decided to salvage what they could of their userbase. I'm seeing an exact repeat (from the view of outsiders) with Vim.
1. When Neovim came along, vim had stagnated. Many features (notably async) in vim 8 probably would not be there if neovim didn't exist.
2. Neovim is "ready". The only feature classic vim has that is really missing from neovim is a really solid GUI (or maybe I just haven't found it yet). And at least for me, the TUI features make running it in a terminal fine.
3. Neovim has a very high level of compatibility with classic vim. Migrating from vim to neovim generally doesn't require changing the config very much if it all. Most plugins work on both. Many patches from vim are carried over to neovim, etc. It's even possible to use the same config for both neovim and classic vim, so you can easily switch between them.
The biggest concern, I think, is that there is some divergence in the APIs for new features, which requires a little bit more work for plugin maintainers to support both. And without any real dependency system for plugins, it's more complicated to distribute a compatibility layer. (and polyfills are impossible since you can't have all-lowercase user-defined function names in vim)
At this point you can mostly think of neovim and classic vim as two implementations of the same thing. Both are innovating, and ideas are shared between them, but sometimes they have slightly different APIs for new or experimental features, but hopefully those apis converge over time.