That is indeed normal for software, but 'beta' testing is also quite a late stage in the testing cycle ... roseway likely tests his software a lot before anyone else even sees it.
For telco-vendor testing, there are usually considerably more test phases than just alpha/beta.
In cases like the ECI rollout, the vendor-customer relationship is ECI-Openreach, not you, the end-user. ECI software developers will run tests before the software is ever included in a product. They will have integration testers running tests (in their labs) to show the whole hardware-software combination works, and interworks with as many 3rd party chipsets as it can, in as many combinations as it can. When a product is to be released to Openreach, a suite of acceptance tests will be run, testing feature combinations that are important to Openreach. When an upgrade is released, more acceptance tests will be performed on the new/changed/fixed features, while selected regression tests will be run on the unchanged features.
It is unlikely that Openreach will be content with ECI's internal testing. They will take the product into their own labs, and run their own acceptance and regression tests, with the hardware that is important to them. They will conduct their own integration tests against their own management and control systems too.
Once signed off against testing in the lab, the upgrade will be allowed onto the live network. It will still be tightly reined in, though - with an initial upgrade done in a friendly environment - perhaps still a DSLAM within Adastral park. More of the acceptance and regression tests will be run with real, live systems in control. In an upgrade like this, those tests will include a wide gamut of modems and DLM settings - especially as it is the way that modems and DSLAMs respond to new methods of interworking with the DLM system.
Once this has shown stability, it will be extended to a few more DSLAMs that are used by the public, and things left to run for a while, with close monitoring. This part is the closest that an end-user would feel to a 'beta' test, but as can be seen, it is a very late stage.
If the change is somewhat controversial, or all-encompasing, I can imagine this phase would slowly grow from the order of ten sites to the order of hundred - a bit like vectoring testing progressed. As the scope increases, the monitoring process swaps from watching how individual lines and modems respond, to monitoring the total statistics. What is seen in the total ES, FEC and CRC statistics? Is quality improving? Is anyone suffering outages?
Once happy with those results, comes the final step: a wide rollout. Monitoring will be statistical, ensuring that the widespread results mirror the final beta tests phase. A rollback at this stage will be a damaging, costly affair ... and all previous testing will be designed to avoid it being necessary.
Don't let @Chrysalis' blunt cynicism fool you into thinking his description is the way things happen.
Something devastating went wrong with the rollout, but it won't be because Openreach only bothered testing with a few random end-users.
Written from the persectve of a software engineer/architect, whose products sit in telco's operational networks, and whose code goes through testing cycles as described, often. It is an unpleasant experience to be hauled in front of a telco when some aspect of a rollout is going wrong; I've had that happen over one or two serious operational problems, but nothing bad enough that a network-wide rollback was necessary.