Simultaneous parallel runway operations at San Francisco International Airport represent one of the more operationally complex traffic environments in the U.S. national airspace system. SFO's 28L/28R runway pair is separated by approximately 750 feet centerline-to-centerline, classifying them as closely spaced parallel runways (CSPR) under FAA standards. When the airport runs departure and arrival traffic simultaneously on these runways — a standard configuration during westerly flow — ATC is operating under the provisions of FAA Order 7110.65, which governs separation standards for exactly these scenarios. The question posed here — what happens if the arriving aircraft executes a go-around while a departing aircraft is already climbing — is not a hypothetical edge case. It is a contingency that SFO tower controllers train for and have written procedures to address.
If the arriving aircraft on 28R initiates a go-around after the departing aircraft on 28L is already airborne, the tower controller's immediate obligation is to issue diverging heading instructions to ensure the two aircraft do not converge. In practice, the departing aircraft would likely receive an immediate turn instruction — typically a left turn off the 28L heading, taking the aircraft away from the 28R go-around track — while the go-around aircraft would be vectored to the right or given a specific missed approach procedure that diverges from the departure corridor. Vertical separation may also come into play: a light to mid-size aircraft departing 28L will often out-climb a heavy widebody executing a go-around, since the go-around aircraft is initially near Vref speed with gear and flaps configured. That performance differential can be an asset to the controller, but it cannot be assumed to provide adequate separation on its own. The controller will work the problem using radar, and in VMC operations, visual separation may be applied as a bridge until radar separation is established.
The pre-V1 abort scenario the poster raises is a separate and more nuanced question. If the arriving aircraft on 28R initiated a go-around before the departing aircraft reached V1, the tower controller would assess whether a stop clearance is appropriate. A rejected takeoff instruction is possible — particularly if runway remaining allows a safe stop and the go-around geometry creates a genuine conflict — but it is not automatic. The controller must weigh the risk of a runway overrun against the risk of continuing into a potential conflict. In many cases, the departing aircraft's own V1 calculation, performance margins, and the available runway length at SFO (approximately 11,870 feet on 28L) would still favor continuing the departure while ATC begins vectoring both aircraft for divergence. Pilots should also be aware that an ATC instruction to stop does not override crew authority; if the crew has passed V1 or determines that stopping is the greater hazard, they retain command authority over that decision.
The broader operational context here reflects a systemic challenge facing high-density airports with legacy runway geometry. SFO's CSPR configuration was designed decades before today's traffic volumes, and it requires a continuous balance between efficiency and safety buffers. The FAA's SOIA (Simultaneous Offset Instrument Approach) procedures, used at SFO during instrument meteorological conditions on 28L/28R, include specific breakout procedures for go-around scenarios and mandate particular missed approach headings precisely to manage this risk. For professional crews operating into or out of SFO, reviewing the ATIS and current runway configuration before descent or departure, briefing the missed approach procedure in full, and understanding the diverging departure heading expectation if flying 28L are all standard practice. What the scenario illustrates is that parallel runway operations are never passive — they require active situational awareness from both the flight deck and the tower, with contingency planning built into every phase of the operation.