Perl 6 will assume that script is in some kind of recognizable Unicode encoding, any of:
UTF-8 UTF-16 UTF-32 SCSU
Of those, probably only SCSU requires a BOM, since Perl scripts are almost certain to be strict ASCII in the first few bytes where it matters.
If it starts parsing as UTF-8, and runs into trouble, it might or might not try to intuit the real encoding. Haven't really decided that yet.
You can always explicitly switch the encoding with "use encoding" or some such.
Is it too late in the Perl 6 process to ask for fewer options here? Saying "it's always UTF8, and if you want it different, you must convert it yourself each time" would save lots and lots (and lots) of problems with guessing. Predictability is good, yes?
