The outcome of one Linx.Sysctl.Reconcile.reconcile/3 pass.

A reconcile pass applies a list of ops (see t:op/0) and records what happened. Sysctl writes are independent per key, so the pass is best-effort: every op is attempted, and any that error land in :failed rather than aborting the rest. (Ordered subsystems like rtnl, where a later op depends on an earlier one, use the same report shape but a fail-fast strategy: :failed holds at most one op and :pending holds the rest. The struct is uniform; the strategy is per subsystem.)

Fields:

• :converged? — true iff nothing was left undone: no ops were needed, or every op applied. false iff anything failed (or, for a fail-fast subsystem, anything is still pending).
• :applied — ops that succeeded this pass, in attempt order.
• :failed — {op, %Linx.Sysctl.Error{}} pairs for ops that errored.
• :pending — ops not attempted. Always [] for sysctl (best-effort attempts everything); non-empty only for fail-fast subsystems.
• :last_applied — the updated ownership map to thread into the next pass (see Linx.Sysctl.Reconcile for its shape and why it is reconciler-held, never persisted). Carrying it here lets a loop do {:ok, report} = reconcile(desired, report.last_applied) each tick.

Inspect

Renders compactly:

#Linx.Sysctl.Reconcile.Report<converged: 3 applied>
#Linx.Sysctl.Reconcile.Report<2 applied, 1 failed>