[evlatests] July 12

[Barry Clark]

It is now pretty clear that letting the Executor sit idle for more
than five or six mimutes is a reliable way of killing Idcaf.  We did
that a total of three times.

Spent time trying to make subarraying work.  Found one bug which was
causing odd behavior when antennas were removed/added to a running
subarray.  This now works as expected (with this not yet committed
code).  However, it does take a while for vispipe/idcaf to get 
reorganized - a two minute scan was entirely lost after each 
remove/add.

When I try running two real subarrays, very strange things happen.
I was running EVLA antennas at X band, VLA antennas at C band.  I was
getting good fringes on EVLA antennas on IFs BD, screwy fringes on
VLA antennas IFs AB, and intermitten, even screwier fringes on the
other combinations.