From @masak

<masak> rn​: say (class {}).^name
<p6eval> niecza v21-1-ga8aa70b​: OUTPUT«ANON␤»
<p6eval> ..rakudo 962b9b​: OUTPUT«<anon>␤»
<masak> rn​: say (class {}) === (class {})
<p6eval> niecza v21-1-ga8aa70b​: OUTPUT«False␤»
<p6eval> ..rakudo 962b9b​: OUTPUT«True␤»
<masak> o.O
<masak> Rakudo! really, "True"?
<masak> r​: say (class { has $.x }) === (class {})
<p6eval> rakudo 962b9b​: OUTPUT«True␤»
<masak> r​: say (class { method foo {} }) === (class {})
<p6eval> rakudo 962b9b​: OUTPUT«True␤»
* masak submits rakudobug

S03 doesn't go into detail about what "value identity" means. But the
consensus intuition seems to be that things of different types should
not compare as True under infix​:<===>, like they do here.

From @niner

Fixed in rakudo commit 86731df15fac5e079ba158bd0b22b93e3cb5ac2c
Test in roast​: Raku/roast@84115c62db

The RT System itself - Status changed from 'new' to 'open'

@niner - Status changed from 'open' to 'resolved'