Have a REST driver doing publisher poll. Works fine, but each night
they "reset" the webservice we are polling. This generates a
"connection refused" failure on next poll and the driver shuts down
with a fatal error.

For other drivers I've been able to catch and manipulate the status
event from for example an "error" to "retry" or "warning".

However this same approach doesn't seem to work here.

The error is:

<nds dtdversion="3.5" ndsversion="8.x">
<product build="20150722_0750" instance="Acme"
version="">Identity Manager REST Driver</product>
<contact>NetIQ Corporation.</contact>
<status level="fatal" type="driver-general">null: An unexpected error
occurred in the publisher channel:Connect to acme.local:8180
[acme.local/] failed: Connection refused: connect</status>

I know from experience with the SOAP driver that one can configure
under driver parameters the HTTP error codes to retry on, however this
doesn't seem to help in this scenario.

The only two solutions I can think of are to use driver health to
restart the driver when it has stopped unexpectedly or use two jobs to
stop and restart the driver once this "nightly maintenance period" has
passed. I don't like either option very much.

Any other ideas?