>
the only limit to your velocity is youI have never seen this to be the case outside small, self-contained projects. Even with a fixed scope (its own can of worms), major slow-downs often clearly come from outside the team—unexpected or unreasonable infrastructure limitations, problems with the products of other teams, shortcomings of internal or external tools and libraries, unclear or inconsistent evaluation of the work by its eventual users... etc.
Moreover, in cultures tied to estimates-as-quotes (or estimates-as-deadlines), there is often significant pressure for "product" teams not to try to fix these problems, even for themselves. It would be net faster for us to write our own version of X than to try to integrate the work of team Y? But they've already done the work, why would you want to do something redundant?
In a world where technical teams have clear ownership, this is less of a problem. Team Y giving you problems? Work around it the best way you can. Infrastructure not supporting your needs? Hack together your own on top of lower-level pieces until it's ready. But organizations with cultures that give technical teams the space to do this also have far less issues with estimates—both in having better estimates and being able to handle estimate uncertainty gracefully.