Half the users still hit the old server, two hours after the migration

• The API was moved to a new host at 14:00; the old host was shut down at 14:30.
• Since 14:30 roughly half of customers report "connection timed out"; the other half are fine.
• From the office, curl https://api.example.com/health returns 200 every time.
• Nothing in the new server’s logs looks wrong — it simply never sees the failing customers.

$ dig +short api.example.com
203.0.113.20                     # office resolver: the NEW address

$ dig +short api.example.com @8.8.8.8
203.0.113.10                     # public resolver: the OLD address

$ dig api.example.com @ns1.example.com
;; ANSWER SECTION:
api.example.com.   86400   IN   A   203.0.113.20

$ dig api.example.com @8.8.8.8 | grep -A1 ANSWER
;; ANSWER SECTION:
api.example.com.   79212   IN   A   203.0.113.10   # 79,212 s of TTL still to go