I'm not really that familiar with the Ethernet2 library, but after a quick look it seems like most of the enhancements apply. From what I see the Client.cpp looks basically unchanged.
My main concerns on the "SPI busyness" have to do with:
-
The impact on the bus if other SPI devices were being used. The tight loops continually checking the status of the socket would monopolize the bus. Maybe not an issue because the master is busy anyway?
-
Potential impact on interrupt handlers. I'm not sure about the interrupt priority of the hardware SPI or if it's even an issue. It just didn't seem like a good idea to be thrashing the bus.
Since a loop is needed to keep checking the status anyway, the only issue is whether a delay(1) serves any value. The only negative is adding a millisecond to the connect or disconnect. There's so much other overhead that I wasn't able to make more than a few connections a second anyway so I can't see 1ms making much difference. Maybe it's not enough delay to help items 1 & 2 anyway and therefore moot?
Anyway, that's the reasoning behind the code.