We had 10 logical flows/nodes setup on our collectors. The default HTTP status port is 11 from the beginning of the range of the collector ports. With the addition of the host/logical mapping, that pushes the (now) 11th mapping into a conflict with the HTTP status port.
The HTTP status port is already claimed so the collector goes one above (35863 in the default setup case). The master failed to see that the RPC port was not bound to 35862 (the next in the series) and pointed all auto chained nodes at 35862 causing log delivery to fail.
The master should check that what it expects on the port is listening on the port in my opinion. It would have been very easy to see that the collector was not listening on 35862.