[evlatests] Software tests July 9

[Barry Clark]

I had previously suggested that the loss of coherence between EVLA and
VLA in the 10 seconds before switching to U band was due to the premature
selection of the alternate sideband in the deformatter FIR filter.
In the interim since then I discovered that in the standard Executor,
that command was sent as immediate execute, rather than tagged as deferred.
This is a sufficient explanation for the decoherence phenomenon.
So I modified my copy of the Executor to send the command deferred.
I unfortunately have a number of other changes in this executor, so it
is not totally clear what is due to what.

Anyway, with the modified Executor, running the script to test this,
it caused Idcaf to die (an unusual occurence in continuum mode).  
Further, it nicely, repeatedly, caused it to die, several times, within
a few minutes of starting the script.  However, as Walter was homing in
on the cause of the problem, it stopped doing it.

Back to the original problem.  It turns out, that with the time deferred
commands, not all commands are being executed by the deformatters.  It
is not clear why.  James had a quick look and eliminated the most likely
network causes.  It is my impression that the standard Executor gets
its (non-time deferred) commands executed a higher percentage of the time 
than my modified one, so I am suspicious of the deformatter time deferral 
mechanism.

However, when things did work right, and the appropriate switches were
thrown, the decoherence problem went away.  But, if commands are missed,
it provides a mechanism for the entire scan following a U band scan to
have an undiagnosed case of no fringes.