This is a universal challenge for all product owners - how to ensure your roadmaps don't get derailed. I take the perspective that you should expect and accept a certain combination of planned vs opportunistic initiatives. Being overly dogmatic in your roadmap against evolving circumstances is just as bad as not having a strategic roadmap.
Amazon is known for moving quickly, but that doesn't necessarily mean that we don't have structured long-term roadmaps. However our culture and mechnisms ensure both a rigor for annual planning while accommodating, and even respecting, the need to change priorities quickly if needed. Our annual OP1 process, usually mid-year for the following year, ensures that we map out our north-star and strategic investments for the year tied to clear business goals. These "big-rock" initiatives may lack the detail of the "how" but they are explicit in the business outcomes we expect and how we measure them. A later OP2 process at the end of the year, does two things - it accommodates smaller tweaks to the original strategy if needs have evolved and also adds more substance to the "how" that will ultimately result in a roadmap. This one-two punch ensures that we maintain the rigor of strategic planning while allowing us to react as market, customer-base and other macro conditions change. Considering our peak-period is Thanksgiving-Christmas, this also allows us to absorb and react to new learnings without being beholden to an older roadmap. Furthermore, we have an ongoing process for conceptual thinking using PRFAQs - essential a concept proposal - that are continuously evaluated that still give us the opportunity for identifying a new opportunity, or addressing a dire need and making prioritization decisions as needed.
In practice, this usually translates to a 2-3 initiatives that usually surface during the year that may not be in the original roadmap. This isn't a defect, it's a necessary balance between the rigor of planning and acting on opportunity. It's a popular adage in finance that you make most of your money on three critical days of the year. It's much the same with product development - feature value is often assymetrical to effort and planning and recongizing this balance is critical for product success.