Removed: over 140 services, each with its own queue and its own repository, and the freedom for those services to run different versions of the 120 dependencies they shared.
Outcome: one service, one repository; the test suite went from up to an hour to milliseconds; shipping a change to a shared library went from 140+ deployments to one; the team made 46 improvements to those libraries the following year, against 32 the year before.
Someone ran the opposite experiment and published the receipts. Twilio Segment’s customer-data pipeline had grown to over 140 services, each with its own queue and its own repository. In their words, “our operational overhead increased linearly with each added destination”, the on-call engineer was “literally losing sleep”, and the test suite “could take up to an hour to run”. They consolidated all of it into one service, one repository, and one agreed version of each of the 120 dependencies. The test suite went from up to an hour to milliseconds, shipping a change to a shared library went from 140+ deployments to one, and the team made 46 improvements to those libraries the following year, against 32 the year before. What they deleted was not just overhead: it was 140 separately patched, separately deployed, separately compromised surfaces.
Source: Twilio, Goodbye Microservices: From 100s of problem children to 1 superstar (10 July 2018)