TTL is a promise to recursive resolvers: you may remember this answer for this long. A 24-hour TTL on a record you might move is a 24-hour outage you pre-authorised.
Not every record wants a low TTL
MX, SPF and stable documentation hostnames can stay high. Application apex, CDN hostnames and failover targets should be low enough to match your RTO. Cloud DNS supports a one-second minimum.
Lower it before the window
Drop TTLs days before a migration, not during it. Otherwise the old high TTL is still in the wild while you are changing records.
Clamping is real
Some resolvers refuse to honour very low TTLs. Health-aware answers still help: even a clamped cache is less likely to point at a dead target if you stopped publishing it.
Query volume is the cost
Low TTL means more queries. That is why Anycast authority exists. A registrar DNS product that discourages low TTLs is often protecting its own capacity, not your RTO.