Engineering Reality.
We get this ticket a lot: page load times are fine on staging, terrible in production, and everyone points at the platform. Usually it isn't the platform.
Open up the request waterfall on a slow checkout and you'll find Magento waiting — on an ERP for stock numbers, on a payment gateway with no timeout, on a shipping API being called once per line item instead of once per cart. Sometimes a PIM sync job has locked the exact tables the storefront needs to read from, and it happened to kick off at 2pm on a Tuesday because nobody scheduled it around actual traffic.
Magento handles all of this fine when you benchmark it alone. Add three or four real integrations, each with its own latency and failure modes, and the math changes completely.
None of it is a bug, really. These are decisions that made sense when the store had less traffic and nobody was watching the gaps between systems. They just don't hold up once the store scales.
What usually helps: cache anything that doesn't need to be live down to the second, put a circuit breaker in front of external calls, batch shipping and pricing requests instead of firing one per item, and move sync jobs to hours nobody's shopping.