Do not confuse phase adjustment, frequency adjustement for channels in the same synchronisation domain, and synchronisation to different sync domains.
Phase adjustment can be very fast, frequency adjustment as well if the two channels are synched to a same time reference.
Imagin two transmitters with two differents frequencies, and one receiver. If the two transmitters are synched to a common time reference, then switching from one channel to the other one in the receiver can be instantaneous, with a synthetised receiver. You don’t need to resync, just ask an new frequency, based on the actual internal time reference.
Internal time reference is extracted from the previous chanel, or received by an external precision clock (atomic time).
Eventually a small phase adjustment is needed at channel switching, but not more.
So for a receiver, if transmitters are synched to a common clock, it’s faster to synchronize when roaming because you can rely on the internal reference, extracted from the previous chanel lock.
This explain why there is a precision local oscillator inside mobile phones, this is to keep sync during roaming.
Syncing to different synchronisation domains is longer to achieve.
When you are roaming between GSM access points, the frequency of each access point is different, but the sync domain is the same, so even if phase and frequency are different seen from the receiver to each transmitter, fast roaming is easy.
Synching with asbolute phase is only needed for transmitters on the same tower, and yes it need a common sync feed. Not really difficult because on the same tower.
The good thing with sync, is that you can use more advanced receivers, with real time roaming, without adding heavy costs on the receiver side.
This was the goal with GSM networks : low cost and realtime roaming receivers.
But I am eager to learn! 