The busiest stretch of track I carry has run out of room — room, not speed. The SBB have started MehrSpur Zürich–Winterthur, the largest project in the federal Ausbauschritt 2035 and the biggest railway construction site in Switzerland. Through 2026 they are preparing the installation areas at the tunnel portals and starting station works at Wallisellen, Dietlikon, Bassersdorf and Winterthur-Töss. From the December 2026 Fahrplanwechsel a few S-Bahn lines already ride altered journeys; SBB say the details land in the app and online timetable from autumn 2026 (SRF, 17 March 2026).

←TODAY: the site works are running and the December 2026 timetable change lands on passengers this winter.
→3012: the corridor that carried the eastern crush without ever going dark is the one the manifesto still runs on.
Fulcrum: you only see why a three-minute tunnel is worth eleven years when you stand at both ends of the century at once.

The centrepiece is the Brüttenertunnel: about 9 km, two single-track bores at roughly 8.3 m inner diameter each, 160 km/h. Nine kilometres at that speed is about three minutes of running time. Eleven years of construction and — in SBB’s own words — “rund 1 Million Franken” a day for a three-minute tunnel reads absurd, right up to the moment you notice the three minutes were never the product. What is: 30 % more capacity — about 900 trains a day for over 150,000 passengers, the quarter-hour as the base Takt, eight minutes off the long-distance run.

Gesamtprojektleiter Bruno Studer, on the record with SRF: “Es sind schnellere Verbindungen möglich, aber auch stabilere. Wir werden weniger Verzögerungen und Verspätungen haben.” — “Faster connections are possible, but also more stable ones. We will have fewer delays and less lateness.” Read the order out loud. Faster comes first because faster sells; stabiler is the thing being bought.

Four tracks where there are now two do not principally make my trains quicker — they decouple them. The fast long-distance service stops inheriting the delay of the stopping service ahead of it, and one perturbation stops propagating across the whole eastern half of the network. On a clock-face timetable, punctuality is bought not with speed but with capacity you deliberately do not fill — the spare path, the second bore, the margin. The Takt keeps time because somebody paid for headroom that sits empty until the morning it doesn’t.

Zürich has run this play before

The Durchmesserlinie took the terminus at the Hauptbahnhof and made it a through-route — four-track Bahnhof Löwenstrasse plus the 4.8 km Weinbergtunnel under the Limmat, in service 15 June 2014, about CHF 2.031 bn at 2005 prices. It deleted the bottleneck. Twelve years on, the binding limit sits twenty kilometres north-east, on the two-track corridor to Winterthur — and buying it out again runs to CHF 2.9 bn in SBB material, CHF 3.3 bn in SRF’s March reckoning with up to 10 % contingency on top.

The same law ran in public on the other side of the planet. Auckland’s City Rail Link opened on 13 September 2026 — 3.45 km, NZ$5.493 bn (Railway Gazette International, 14 September 2026) — turning the Britomart dead-end into the through-station Waitematā, the exact operation Zürich ran in 2014. Three days in, the thing now rationing frequency is level crossings on the Western Line, far from the new tunnel: Auckland Transport runs a six-month transitional timetable at six trains an hour rather than the eight tested (1News, March 2026). Two cities, one property of every network: remove the binding constraint and you promote the next one. Nobody mis-built anything in Zürich or Auckland — the honest question at a megaproject’s start is not what it unblocks but where the constraint moves next, who owns that one, and whether it sits in the same business case.

Atelier: Nothing closes. My corridor carries the country’s densest traffic all day, so the heavy work happens at night, between the last train and the first. Set the two bridges against that clock — 580 m of single track at Wallisellen (2026→2031, five years), 800 m at Neumühle (2027→2034, seven years). No engineer needs five years for 580 metres; the programme is set by the hours a year you are allowed on the site, not by the span. On any job built under occupation — a school in term, a hospital wing, a listed concourse — two-thirds of a night shift vanishes into mobilisation, isolation and clearing the track before the first service. Your Monday move: put the access regime — possession windows, isolation times, the hour the site must be clear — into the first programme diagram, next to the site survey.

Hack: Size the programme by usable hours and watch the night-window overhead do the damage. The parameters below are illustrative, not SBB figures.

window, setup, teardown = 5.0, 1.0, 0.75   # last train -> first; illustrative only
usable = window - setup - teardown
nights, site_hours = 180, 9000
print(f"{usable/window:.0%} of the window is work -> {site_hours/(usable*nights):.1f} years")

Cut the scope by a third and the answer barely twitches; widen the window by an hour and it collapses — the whole economics of building under Betrieb in one print statement.

Credit where it carries load: SBB as client and operator, Implenia on the Brüttenertunnel, AFRY on the engineering — up to 1,000 people across a 30 km perimeter, for a decade, without ever closing the busiest railway in Switzerland. As PAZ’s own concept work on infrastructure systems puts it, the interface between subsystems is the object of design. Ask the megaproject question of your own project: when your bottleneck clears, where does the next one appear — and is it inside your scope, or somebody else’s? If it’s somebody else’s, say so on paper now, while it is still cheap.