Also that has nothing to do with thread::interrupt. In such case
valgrind-clean exit may happen only after blocking call finishes. So
it does not illustrate flows in thread::interrupt, more like "example
of mt programming caveats".

In no way i encourage pointless arguing (my problem is long solved)
and i am sure you know better than me what thread::interrupt can and
can not do. What to Dale his comment was emotionally-driven and mainly
revolved around choosing right epithets and so. Most likely it just
makes no sense.