It only affects about 1% or less of the entire client base (roughly <30 phones total of over 3900+ users).
The cluster had been fine for about 3 years, and in the last two months something has this handful of phones not coming up (without a physical reboot/unplug per phone AFTER the reboot).
Interesting.
Can you check your controller's "System Capacity" form?
"Configurable IP Users and Devices" = ?