Well state and primary-variable handling: direction for the open well PRs

[hnil]

There are currently 14 open well-related PRs from me, ten of them drafts, some dating to 2022. Reviewed one at a time they look like unrelated small fixes, which is why most have stalled. Together they are one question, and I would rather settle that question than keep pushing individual patches.

The question: should well primary-variable state be owned and persisted by the simulator, or always recomputable from rates?

These four all assume the former, and all still apply to master:

• Safer well state #5990 — persist primary variables in SingleWellState, including in serialization, so they survive restart rather than being rebuilt from rates
• added safer but potentially extra calls to update primaryvariables #5972 — assert that stored primary variables match the well state
• avoid recreating wellcontainer including deleting informations form p… #5991 — stop clearing/rebuilding well_container_ unconditionally, so per-well cached data is not discarded
• Well primaryvariable scaling #4176 — scaling of well primary variables (cannot be rebased as-is; kept as a marker)

Today's code assumes the latter. That disagreement, not the code, is why these sit.

Second question: does #7185 (the experimental well model) supersede this line of work? If it is the direction, I will close all four. If it is a parallel experiment, they should be judged on their own merits.

Unrelated but overdue: #7020 has been open, non-draft and mergeable since April with no review. It either needs a reviewer or a decision to close.