>>It's funny, though, that it occasionally gave "illegal operation", but then did read all the memos in the scan-endscan loop. Maybe the weak spot was on a deleted record's memo?
>
>That's a good point and I'd be interested to know. Perhaps George can tell us? (I'll copy this to him)

Hi David,

What happens is that the pointers and other information regarding the memo file are re-built by SCAN...ENDSCAN, REPLACE ALL methods, and (I think) even a straight PACK. There are a number of different reasons why this information can become corrupted, but in some cases, if the damage isn't severe, it can be corrected using these techniques.