Hourly Tracking Runs in Its Slot

Monitors that track every hour, every 6 hours, or every 12 hours now get each run inside its time slot. A run that comes in late still counts toward the slot it belongs to.

• Runs for these monitors go ahead of daily tracking, current slot first.
• A run that misses its slot moves to the front of the queue on the same model, and its answer is recorded in the slot it was meant for.
• ChatGPT and Perplexity keep capacity free for these runs in every country, at night too.
• You can't pick a frequency your remaining response quota can't cover, counting what your other monitors will use. The same check runs when you add models to a monitor or turn it back on. Lowering a frequency always works.
• In the REST API, those requests return a 402 with the MONITOR_PROMPT_FREQUENCY_OVER_QUOTA error code.