It is possible that an entry in the search database is not
pointing to the correct location.
But the problem is server response time, I think.
RH divides the search chores by using a multilevel database.
Keywords and pointers are grouped alphabetically into several
files. They reside in the whdata, whgdata and whxdata folders. A
master file says, for example, that Aardvark to Donkey is in the
first one, Dumbo to Fox in the second, and so on.
The search looks first in the master file and consults only
the relevant file, which makes it faster than starting with A.
However (I'm guessing here), this depends on all the files being
resident on the PC. When the search goes looking for a hunk of the
database that isn't loaded, you get the error.
You can test this. When you search for something beginning
with "A", does it work better than for a keyword starting with "T"?
Does a search for a "G" word sometimes work, and sometimes
not?
Same for the index, except the index starts to load before
you enter the term. If it's not completely loaded, it works for A
but maybe not for Xylophone.
This brings me to my favorite pet peeve.
Disclaimer: I am not an expert in the following subject, but
have a strong hunch. Feel free to disregard or correct me.
The number of impediments to server response time is growing.
Let's assume hardware and transmission lines are not to blame.
Programs that run network traffic cops and source authentication
now look into some places where they previously did not, I believe.
Maybe at the packet level?
When the highway's smooth and the weather's clear for a
caravan, how many tollbooths are ahead?
As I say, I hope some networking/security expert will explain
why I'm wrong.
Harvey