Caisson wrote:

> Normally I would discard such a story as an hoax.

  No hoax.  Well-known aspect of IC design.

> What's so specific of those internal protection-diodes that they can't
> be used like their external counterparts ?

  Because they aren't external!  They are in fact on the same silicon
die as all the other components.  Recall that a transistor is in fact
just two diodes.  But it doesn't *behave* as just two diodes, as charge
injected into the junction of one diode can be passed into the other.

  Well, these on-chip diodes are similar.  They are sitting near other
components, other junctions on the die, and if you inject *enough*
current into them, they may firstly cause other (reverse-biassed)
junctions to conduct.

  Worse, other junctions may also inject charge back into them.  If both
things happen together, there is an avalanche effect which happens to be
the same as in a thyristor, and the whole chip may "switch on" from rail
to rail.  Nasty!

  Common occurrence in CMOS chips such as 4066s which are *not*
symmetrical.  Connected one way, they always behave, but swap the
"switch" connections over and pull too much current and - very hot chip!
--
  Cheers,
        Paul B.