There are 16 jobs per molecule. If you do 3 or more jobs correctly, that is useful. Once the algorithm fails, the task is over and returns the job results that you did.

Wingman _9 of the WU that I reported above has done a similar trick. (E214922_ 872_ A.35.C29H17NS3SeSi.38.4.set1d06)
He's Pending Validation but got RC = 0xc0000005 after job 12, while I got Error with the same job log as far as I can see.
I agree with Apis Tintinnambulator - what's going on?

Hi Lawrence,

Again, thanks for looking into this.

I rather assumed that there must be some information in the returned files that is not reflected in what we see in the logs and it was just an annoying "feechure". But other than to say that I hope it will be made clearer in future science apps, I'm content enough just to live with a minor annoyance in this one (so long as its frequency doesn't increase). If many others report it, however, I think it would justify a little closer attention from the techs, even if only to keep us crunchers happy: I for one certainly like to know why my machines are producing bad results.

I just had a look at the Results Status here @ WCG.org for 2 CEP2 WUs in my cache on this device that have short Deadlines and are still Ready to Start.
They have both returned Error status to all other wingmen.
Most got RC = 0x100, a few RC = 0x01, but none RC = 0xc0000005
A few qinky ones too.

E214649_ 089_ C.35.C31H16S3Se.00985909.1.set1d06_ 4--
E214903_ 644_ C.36.C33H20N2S.00932953.3.set1d06_ 2--

I intend to periodically go through all of "my" CEP2 WUs that haven't yet been returned, and abort any for which about 3 or 4 or more wingmen have already returned an Error. That may* advance them one step closer to the system ceasing to send out repair copies, without wasting my CPU time ...

There must be something indeterminate in the science program's data space that is causing different results on different machines with the same input filles.
It could be dependent on the CPU hardware, and the BOINC sytem may or may not return enough CPU details to the server to be able to discern this.
Or if the CEP2 program does not explicitly initialise one or more of its variables, it could be reading the "random" data that was in the RAM before that chunk of memory was allocated to that CEP2 process.
Or the program may explicitly generate and use random numbers, eg for Monte Carlo simulations, and these sometimes send things out of bounds.
Or, ...

Roll on the rumoured * sparkling all-new improved * version of the Q-CHEM/CEP2 program!
(Any ideas on its time of arrival?)

[Edit:] * Does "Aborted by User" count towards the max no of repair copies of a WU that are sent out? [/Edit]