Curiouser and curiouser...
I followed your steps, as annotated below, and found basically exactly what you did. Could the content of the VI affect this?!?
-Verified that no files existed in ni-rt\startup on the RT Engine.
I don't normally do this, but did this time.
-Targeted LabVIEW Real-Time Development System to the RT Engine
-Wrote a VI, test1.vi, that just writes the character "a" to file called c:\ni-rt\startup\temp.txt.
-Tools>Build application>Source files>Add top-level VI (test1.vi), did not include error files>build>close VIs>OK>Done>No
I normally do not explicitly delete the error files, but as below, this made no difference
-dir in FTP shows startup.exe created at 17:26.
-Tools>Network Options>RT Target: Misc>Launch Application at Bootup checked.
How? This comes back to my confusion about the window visible at this time: As soon as I close all VI's, I return to the dialog you described this morning, and have no Tools or other menus.
-Close LabVIEW, reboot RT Engine.
I don't normally close LV. Instead, I Target to the host PC, before rebooting the PXI. I tried closing LV this time, but in the later retries, I just changed Target. No difference found.
-Using DOS FTP, retreived temp.txt and saw "a". Deleted temp.txt.
-Target Development System to RT Engine. Closed all VIs and updated. Opened test1.vi.
-Tools>Build application>Source files>Add top-level VI (test1.vi), did not include error files>build>close VIs>OK>Done>No
-Use FTP to list directory contents of ni-rt\startup.
-The new startup.exe was created at 17:35.
Other than the different time stamp, I agreed completely.
I continued to retry. Another Build, and I had a new time. Build, without rebooting the PXI first, still new time. Build without deleting the error files (from PDI Control Toolset; why are these even there?), still new time. Usually a slight variation in byte count, by the way. Not sure if that is significant. (130323, 130315, 130311, 130315, 130329...)
I now turned to Building and testing my own GPIB VI...
Built, checked time stamp, Built again, new time stamp, Built once again, new time stamp. Beginning to think I was imagining the whole thing!
Note that I had not been rebooting the PXI between those Builds. (Had not in the latter parts of re-testing your VI, either.) I now rebooted the PXI, and verified the VI ran on boot. Recovered control, Targeted the PXI, and Built. ftp'ed in, and found OLD time stamp.
Without rebooting, switched Target to PXI, Built, checked time, still OLD time stamp.
Going to go home...