Building a homelab Kubernetes platform on Talos (1→N nodes)

## Spark

An eleven-issue epic dropped in one window laying out a Talos Kubernetes homelab from bare mini-PC to scale-out: bootstrap Talos + Flux via OpenTofu, democratic-csi against TrueNAS, Plex/QuickSync migration, FoundryVTT singletons, an NFS worktree/watcher spike, a stateless HPA/PDB tier, and 1→N scale-out.

## Why it could be interesting

The claim: a single-node cluster designed day-one for N-node scale-out costs little upfront and dodges the rebuild — the real skill is choosing which HA/scale settings to bake before you need them (a reversibility-filter decision, cf #205). The most personal angle is the NFS worktree/watcher spike (#247): does my git-worktree daily driver (#254) survive on network storage?

## Open questions

- [ ] Do worktrees + file watchers behave on democratic-csi/TrueNAS NFS, or does it force local-disk scratch?
- [ ] Which day-one Talos machineconfig settings are worth baking vs deferring?
- [ ] Is this the build-narrative counterpart to documenting the homelab as a system (#251)?

## Source material

- alunduil/alunduil-infrastructure: #241 (parent), #242–#250
- Related: #251 (documenting the homelab), #254 (worktrees as daily driver), #205 (reversibility filter)