NB: Prior to this refactoring (prior to 7.0), Index Read would open parts during query initialization time.
The code was written to lazy open on demand, but calls through to the key manager(s) provoked early open.

That was generally not a big problem, except when indexes were being read in child graphs.
In those cases the CQ had access to all parts of the index and the initialization could take some time.
In extreme cases where a CQ had lots of indexread and perhaps those indexes were superfiles, the total time the CQ took to initialize exceeded an internal timeout and the query would fail. The error seen under those circumstances was:

A workaround for this issue, is to increase the timeout involved (default is 5 minutes)
This timeout can be increased with, e.g.:

.. which would represent 15 minutes.