We also ran into one problem with the onboard Wi-Fi that only occurred whilst running the new board in headless mode. This ended up being caused by a power-saving mode on the connectivity chip. Effectively the board became unresponsive in headless mode if it was connected to the network only via Wi-Fi.
“We’ll be shipping a patch at some point to give people the option to disable the power saving.” — Eben Upton
In the meantime, if you’re running the board headless remember to enable keep-alive in your ssh client or you’ll run into a similar problems.
This sounds exactly like what is happening to me - is this keep-alive thing implemented already, or is it something I can try?