Well, I think the answer lies within Reach’s wpa_supplicant log (journalctl -u wpa_supplicant --no-pager).
Overall, my guess is that for some reason the hostapd starts advertising the SSID and all, but rejects Reach’s incoming connection and wpa_supplicant does not attempt another one. I might be wrong, though.
If I am right, the best thing to do is to optimise hostapd startup time.