In https://hackage.haskell.org/package/base-4.19.1.0/docs/Control-Exception.html#interruptible, we have the guarantee that
STM transactions that do not use retry [..] are guaranteed not to be interruptible
But what about the converse - I have an STM operation using retry that I desire to be interruptible (at least when it encounters the retry).
So, strictly reading the guarantee, it is allowed for GHC to keep that retry-using STM action non-interruptible. But does it?
(Edited: I mis-attributed some non-delivery of an interrupt due to the retry-ing STM action not getting interrupted, when in fact I had an earlier leaky interruption point that got interrupted. So in the end, the retry-ing (and orElse-less, see comment from @Taren) STM action was interruptible as expected).