Deploy and manage solutions
Package everything as a solution
A deployable unit contains the cloud flows, desktop flows, connection references, environment variables and any Dataverse components. Desktop flows that belong to a solution are stored in Dataverse, which is why a solution-aware approach is required for pipelines and managed deployment. Choose a publisher prefix early; renaming later is painful.
Managed vs unmanaged
Unmanaged solutions are for development and for transports into environments you keep editing. Managed solutions are for test and production: they are locked, support clean upgrades, and can be removed without leaving orphaned components. Importing an unmanaged solution into production leaves components that remain after uninstall - a common exam trap.
Environment variables and connection references
The default value is a placeholder that ships with the solution; the current value is the environment-specific setting. Current values are not carried automatically between environments unless you supply a deployment settings file, so record them per environment. Connection references must be mapped to real connections at import time, otherwise flows arrive broken. Owners should be assigned after import so scheduled runs do not fail on a disabled account.
Pipelines and automation
Power Platform pipelines, or the Build Tools / GitHub Actions style approach with a service principal, move solutions between environments on a repeatable cadence. Include a pre-deployment check that the target environment's DLP policy permits every connector used.
Worked example
A desktop flow works in Dev but fails after import to Test. Cause: the connection reference was never mapped and the Key Vault environment variable still holds the Dev vault name. Fix: provide a deployment settings file and map connections during import.
Rule of thumb: one solution per business process, one settings file per environment.