When I worked at NetApp we had customers in both sides of the world, what we discovered was that the 'enterprise' guys often had a cadence, it went something like grow, identify challenges, coast, upgrade, repeat. That cycle was built around taking the latest release of 'big player' software, qualifying it, then putting it into production, growing a bit, figuring out where the problems were, talking with their vendor, waiting, getting a new release and then repeating the cycle. But the cadence at 'web' players was sporadic at best and often frenetic, feature request, test, iterate, feature, iterate, test, coast, coast, coast, feature, feature, feature, test, coast, Etc. As ideas struck the 'web' infrastructure folks they would would immediately implement some sort of prototype or test case, then rapid fire come up with features the infrastructure needed to support to maximize this feature. But sometimes they would go long periods, months, where nothing would change (in the infrastructure side, web pages, UX, etc sure but same servers and storage assets).
What I learned from that was that while it was great to have some other company own the burden of the 'base infrastructure' in terms of operational expense (hey you don't need Linux hackers if your not changing stuff at that level, so its a savings in staff etc etc) it imposes a hard limit on the change derivative. What happens if you try to change faster than the infrastructure can change, is that you end up hacking around the limits, and that builds up technical scar tissue that over time slows your mobility still further.
So bottom line, you can't continue to pivot faster than your infrastructure, if you're hacking your way around the infrastructure to change, then your ability to change will die by a thousand hacks. If you find your self thinking you need to hack around your infrastructure, listen to that warning and start planning for a more agile base to work from now rather than when we're struggling to keep an old system working while developing a completely new one.