I understand the frustration of manufacturers dealing with the 3rd edition, but we have to take care of blaming the test labs.
The core problem is with the standard, which was written from a regulatory perspective (in particular, EU regulations), not for test labs. Given the scope of the standard, it is little wonder labs are not charging US$250,000 or more. For areas such as risk management, essential performance and programmable systems, the standard is poorly written and test labs are left struggling to even keep staff who are willing to deal with it.
Ideally the standard would have been split into two sections: part A which is written with clear, testable objective specifications, and Part B dealing with subjective areas like risk management, programmable systems, usability and so on. Part A is deliberately written in a way that fits within the framework of laboratory systems (ISO 17025) and product certification (Guide 65), with Part B written using management systems, intended for manufacturers to implement under their own responsibility, and subject to regulatory audit under the framework that is normally used for regulation (e.g. Guide 62, ISO 17021, 19011).
There may also be some interface between the two (e.g. specifications of Part A can be the output of Part B) but the important point is that the test lab is not responsible for the content of the specification, only the accuracy of the tests according to the specification.
Product certification schemes like NRTL/SCC marks should primarily work in Part A, and if any of Part B is required, there should be boundaries which allow for practical assessments, e.g. NRTL limits to basic safety (electric shock, mechanical injury, fire, thermal etc) which are incidental to the clinical purpose.
The CB scheme must be limited to Part A since it does not have the structure to handle Part B, and it is unable to provide value for money nor improve safety. In fact, the attempt to handle Part B under the CB scheme at the moment is reducing safety, as it forces test labs to write reports for things they are not qualified to handle.
In the context of regulations, a manufacturer should be able to voluntarily select what parts of the compliance are assessed by third parties.
Part A should be structured so that manufacturers can get value for money by using a third party test lab, i.e. it should cost less to do in an external lab than it would to do in house. The external lab has the equipment and experience which allows efficient service. A budget of US$10,000 should be plenty to evaluate Part A for most devices.
Part B could be subject to voluntary audit, but any company that provides the audit should be genuinely qualified in technology of the product, and be able to take partial responsibility for the results (to the extent of the audit).
Of course, all of this is just a pipe dream