[evlatests] Subarrays

[Barry Clark]

Subarrays worked a little better today.  Not entirely clear why.  Possibly
that I did a little general cleanup, including not sending a garbage 
delay record at Executor startup.  Unknown what effect that should have
on subarrays, since it is sent only once when a new Executor is started,
before any scripts are run at all.

I started two continuum subarrays, and this worked nominally.  (Just for
reference, I've seen starting sub2 first, and have never gotten around
to trying to start sub1 first in any case.)

I started two line subarrays, doing fairly prosaic things, which also 
worked nominally.

I started two line subarrays, with the controlling subarray changing 
mode at regular intervals (1A -> 1B -> 1C -> 1D -> 1A etc).  Subarray
1 behaved nominally.  Subarray 2 did not register a new scan start when
the mode changed (I didn't think to check whether the right mode was 
appearing in the data record).

I killed the script in subarray2, and restarted it.  At this point things
became very weird.  Subarray 1 had good data, except that, although
the Executor and Idcaf believed that modes were cycling as above, the
correlator did not, so mislabeled data was being written.  Subarray2 had 
garbage data of the worst sort.