Possible regression in mocking

[nohwnd]

dataplat/dbatools#10405 (comment)

[nohwnd]

Closing as by-design.

This is the intended behavior introduced in #2654 (fixing #2178): a Mock with a -ParameterFilter that doesn't match a call — and with no default (unfiltered) mock to fall back to — now throws instead of silently invoking the real command. The silent fall-through masked real bugs, so the new behavior is deliberate and the message is actionable ("none of the parameter filters matched, and there is no default mock to fall back to").

The dbatools report is actually a good example of why the throw helps: Install-DbaInstance.Tests.ps1 mocks Get-DbaCmObject only for Win32_processor, but the code also calls it for win32_ComputerSystem (via Resolve-DbaNetworkName). Under Pester 5 that silently hit the real CIM call; under Pester 6 it throws — except dbatools swallows the error in a broad catch, which is what produced the confusing downstream WinRM failure.

The fix is on the dbatools side (add a default/broader mock, or mock Resolve-DbaNetworkName). Full root-cause writeup and a suggested fix: dataplat/dbatools#10405 (comment dataplat/dbatools#10405 (comment)).

If we later decide the "mock for specific args, real command for the rest" pattern deserves a first-class opt-in, that's a separate feature request rather than a regression.