Now I'm utterly confused, and I've no longer the faintest clue what you're trying to say.

Ask a twisted question, get a twisted answer.

you seem to disagree with me on 2).

Nope.

But that doesn't actually seem to work:

Not for regex, unfortunately, but it does for uc

$ perl -CS -wE'no  feature qw( unicode_strings ); $_="\xEB"; utf8::dow
+ngrade($_); say /\w/||0; say uc;'
0
ë

$ perl -CS -wE'no  feature qw( unicode_strings ); $_="\xEB"; utf8::upg
+rade($_);   say /\w/||0; say uc;'
1
Ë

$ perl -CS -wE'use feature qw( unicode_strings ); $_="\xEB"; utf8::dow
+ngrade($_); say /\w/||0; say uc;'
0
Ë

$ perl -CS -wE'use feature qw( unicode_strings ); $_="\xEB"; utf8::upg
+rade($_);   say /\w/||0; say uc;'
1
Ë
[download]

I think you had UTF-8 source code without using use utf8;.

I was thinking about Perl 5.12 implementing Unicode 5.2 (as opposed to Unicode 5.0.0 in 5.10).

Yes, you get that for free.

$ corelist -a Unicode

Unicode was first released with perl v5.6.2
  v5.6.2     3.0.1     
  v5.8.0     3.2.0     
  v5.8.1     4.0.0     
  v5.8.2     4.0.0     
  v5.8.3     4.0.0     
  v5.8.4     4.0.1     
  v5.8.5     4.0.1     
  v5.8.6     4.0.1     
  v5.8.7     4.1.0     
  v5.8.8     4.1.0     
  v5.8.9     5.1.0     
  v5.9.0     4.0.0     
  v5.9.1     4.0.0     
  v5.9.2     4.0.1     
  v5.9.3     4.1.0     
  v5.9.4     4.1.0     
  v5.9.5     5.0.0     
  v5.10.0    5.0.0     
  v5.10.1    5.1.0     
  v5.11.0    5.1.0     
  v5.11.1    5.1.0     
  v5.11.2    5.1.0     
  v5.11.3    5.2.0     
  v5.11.4    5.2.0     
  v5.11.5    5.2.0     
  v5.12.0    5.2.0     
  v5.12.1    5.2.0     
  v5.12.2    5.2.0     
  v5.13.0    5.2.0     
  v5.13.1    5.2.0     
  v5.13.2    5.2.0     
  v5.13.3    5.2.0     
  v5.13.4    5.2.0
[download]