Perl generally can't be fit into that model, since creating and tearing down interps will take too many milliseconds per callback run.

You read too quickly.

My solution is to never try to move the Perl thread off its original OS thread.

That implies that the OP -- not me; I didn't construct the problem, just help with the solution -- had a choice. I contend, very strongly -- that he did not.

In threads, what defines a thread is a 'context', usually abbreviated CXT, and prefixed with xxx_, Eg. MY_CXT; which in turn is just an alias for typedef struct _ithread. The only thing that tie a particular ithread, to a particular OS thread, is that the context (pointer to struct _ithread) is stored in a thread-local storage.

When using asynchronous IO, the user code has no control over what OSthread will be running when the OS callback to complete the IO.

And if the 'wrong' OS thread -- ie. *not* the one that created the IO completion callback -- is running when the IO completion callback is called by the OS, then things go belly up (Ie. CRASH!), because none of the perl environment elements -- stash, coderefs, closures, et al. -- are available when the Perl coderef given as the callback, are available if it is called within the context of the wrong thread.

The *only* way to address that problem is to set the context for the current thread -- at the XS level -- prior to entering the Perl level callback code.

If that description does not seem correct and unassailable to you, it can only mean that you do not understand the problems inherent is responding to asynchronous callback at the Perl level. So read more and think twice.