Crash test update

[Stephan Bergmann]

On 10/08/2014 09:53 PM, Miklos Vajna wrote:
> I *think* the assert is correct and the RTF tokenizer has to be
> adjusted, but first I wanted to see how much does
> e1cbaebe7fe36690e192778c87b5eb63790017d7 help. In any case, it's good to
> see that this time the jump is around 5 documents and not 1000. :)

Just curious, and as those "Crash test update" mails with their obscure 
.csv attachments are white noise to me---how do you read from that data 
that there was a jump of 5 in bad documents, for example?

Stephan