Recommended by whom? Microsoft Corp.?

Not sure what Microsoft's official recommendation is in this regard (if anyone knows, please share). My "is recommended" statement is just my resumé from personal experience, in particular from having worked in Japanese Windows environments for a couple of months.

My impression there was that overall you'll run into the least problems if you always tag unicode files as such using a BOM (be they UTF-8, UTF-16 or UCS-2). Some programs will try auto-detection (with varying success), but many simply assume the file is in the default legacy encoding, if not told otherwise.  YMMV of course, depending on which applications you're primarily working with. So please take this with a grain of salt.

I don't like BOMs in UTF-8 files on any platform...

I personally don't like them either, in particular on Unix platforms, where they tend to create more problems than they solve. OTOH, I've gotten used to the situation that different platforms have different approaches and philosophies.  After all, with Perl in my handbag, this isn't too much of an issue anyway...