From @smls

  say <a b> X~ (1, 2, 3); # (a1 a2 a3 b1 b2 b3)
  say <a b> X~ |(1, 2, 3); # (a123 b123)
  say <a b> X~ (1, 2, 3).Slip; # (a123 b123)

Those last two seem wrong. It's as if the meta-operator thinks it got
four arguments, i.e. as if it was called like​:

  say &[X~]( <a b>, 1, 2, 3 ); # (a123 b123)

I think in keeping with the single-argument rule, the X and Z
meta-operators should behave as if both sides were bound to a +@​
parameter, which would "disarm" the Slip and keep it from messing with
the internal algorithm​:

  sub a (+@​a) { dd @​a };
  a |(1, 2, 3); # [1, 2, 3]