feat(memory): enable zram compressed-RAM swap by default (#76 slice)

First slice of BACKLOG #76 (hibernation + zram, split): zramSwap on by
default in modules/nixos/oom.nix — the memory-pressure layer before
earlyoom. zstd @ 50% RAM (nixpkgs defaults, explicit for reproducibility),
priority 100 so day-to-day paging fills zram first and any disk swapfile
stays reserved for #76's hibernate image (hibernation can't resume from
volatile zram). Follows the earlyoom mkDefault precedent — no toggle.

Adds checks.zram-swap runNixOSTest importing oom.nix, asserting the device
is an active swap, uses [zstd], and carries priority 100.

Verified: V2 checks.zram-swap PASS (/dev/zram0 in /proc/swaps, zstd,
prio 100); V1 full distro toplevel builds; V0 flake check --no-build green.
#76 remainder (disk swapfile default + resume + installer) still open.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

agent/BACKLOG.md

@@ -28,12 +28,18 @@ in [`docs/ROADMAP.md`](../docs/ROADMAP.md); map in
 
 ## NEXT
 
-### 76. Hibernation default + zram `[big]`
+### 76. Hibernation default `[big]`
 **Product intent (Bernardo 2026-07-10):** hibernation on by default;
 zram on alongside for live memory pressure. Disk swap ≥ RAM for the
 hibernate image; zram high-priority for day-to-day swap (not sufficient
 alone for resume).
 
+**✓ zram slice shipped (2026-07-10):** `zramSwap` on by default in
+`modules/nixos/oom.nix` (zstd, 50% RAM, priority 100 so day-to-day paging
+stays off the disk swap reserved for the hibernate image); `checks.zram-swap`
+VM test asserts device/algorithm/priority (V2 pass). **Remainder below is the
+hibernation half** — disk swapfile default + resume + installer.
+
 **Reference layout (this machine — already hibernation-capable, no zram
 yet):** LUKS (`crypto_LUKS` → mapper `crypted`) holds one BTRFS with
 `@swap` mounted at `/swap`; swapfile `/swap/swapfile` (here 16G);
@@ -47,12 +53,12 @@ Installer path already sketches this (`@swap` + resume_offset patch in
 
 **Scope (split before/while implementing — do not land as one mega-PR):**
 
-1. **Design** — default `swapSize` (= RAM? round up?), zram size/priority
-   vs disk prio, LUKS+systemd-initrd resume (already used), what when
-   install chose `swapSize=0`, migration for existing machines.
-2. **Module** — NixOS: enable zram by default; ensure hibernation knobs
-   are coherent when swapfile exists (resume device/offset); Power menu
-   Hibernate already present — must work when swap is configured.
+1. **Design** — default `swapSize` (= RAM? round up?), LUKS+systemd-initrd
+   resume (already used), what when install chose `swapSize=0`, migration
+   for existing machines. (zram priority-vs-disk already settled + shipped.)
+2. **Module** — NixOS: ensure hibernation knobs are coherent when swapfile
+   exists (resume device/offset); Power menu Hibernate already present —
+   must work when swap is configured. (zram already default-on in oom.nix.)
 3. **Installer** — default new installs to hibernation-ready `@swap`
    swapfile (encrypted with root); keep `0` = no swap opt-out; document
    migration.
@@ -62,11 +68,9 @@ Installer path already sketches this (`@swap` + resume_offset patch in
      (`runNixOSTest` or install-test style) — encrypted root + BTRFS
      `@swap` swapfile + resume_offset; prove hibernate→resume path at
      least as far as QEMU allows (fail the item if only “config
-     evaluates”). Also assert zram is active with expected priority
-     relative to the swapfile.
+     evaluates”).
    - **V3:** real laptop (this LUKS+@swap layout): Hibernate → power off
-     → resume session; queue HARDWARE-QUEUE. zram under load is
-     secondary.
+     → resume session; queue HARDWARE-QUEUE.
 
 **Out of scope:** TLP; formatter. Prefer boring NixOS knobs; extend the
 existing installer swapfile story rather than inventing a second layout.

agent/JOURNAL.md

@@ -17,6 +17,22 @@ Template:
 
 ---
 
+## 2026-07-10 — #76 zram slice: compressed-RAM swap default
+- **Task:** BACKLOG #76 (`[big]`) — split; took the low-risk zram-default slice.
+- **Did:** `zramSwap` on by default in `modules/nixos/oom.nix` (zstd, 50% RAM,
+  priority 100) as the memory-pressure layer before earlyoom, following the
+  earlyoom mkDefault precedent (no toggle). High priority keeps day-to-day
+  paging in zram so a disk swapfile stays reserved for #76's hibernate image.
+  Added `checks.zram-swap` runNixOSTest (imports oom.nix). Split #76 in
+  BACKLOG → remainder is the hibernation half (swapfile/resume/installer).
+- **Verified:** **V2** — `nix build .#checks…zram-swap` PASS: /dev/zram0 active
+  in /proc/swaps, `[zstd]` comp_algorithm, priority 100. **V1** full distro
+  toplevel builds. **V0** flake check --no-build green.
+- **Pending:** hibernation half of #76 still open (needs hibernate→resume VM
+  test + V3 laptop). No V3 for zram (under-load tuning is secondary).
+- **Next suggestion:** #76 hibernation half — design default swapSize + the
+  resume/installer wiring; it's the boot-critical, V2-hibernate + V3 part.
+
 ## 2026-07-10 — #41 hardware-confirmed: OpenFile portal floats (V3 close)
 - **Task:** BACKLOG #41 — final on-hardware confirmation (Bernardo reported PASS).
 - **Did:** #41 already out of NEXT + portal rule shipped

flake.nix

@@ -659,6 +659,31 @@
             '';
           };
 
+          # zram compressed-RAM swap (oom.nix, BACKLOG #76 slice): enabled
+          # by default as the memory-pressure layer *before* earlyoom. Prove
+          # the device comes up as an active swap, uses zstd, and carries the
+          # high priority that keeps day-to-day paging off any disk swapfile
+          # (which #76 reserves for the hibernate image). Imports oom.nix
+          # directly so the check tracks the module, not a hand-copy.
+          zram-swap = pkgs.testers.runNixOSTest {
+            name = "nomarchy-zram-swap";
+            nodes.machine = { ... }: {
+              imports = [ ./modules/nixos/oom.nix ];
+              virtualisation.memorySize = 1024;
+            };
+            testScript = ''
+              # zram-generator brings the device up as swap shortly after boot.
+              machine.wait_until_succeeds("grep -q /dev/zram0 /proc/swaps", timeout=60)
+              # Configured compression algorithm is the selected one ([zstd]).
+              machine.succeed("grep -q '\\[zstd\\]' /sys/block/zram0/comp_algorithm")
+              # High priority (100) so the kernel fills zram before any disk
+              # swap — the ordering #76's hibernate image depends on.
+              machine.succeed(
+                  "awk '$1==\"/dev/zram0\" && $5==100 {ok=1} END{exit !ok}' /proc/swaps"
+              )
+            '';
+          };
+
           # Snapshot GUI canary: btrfs-assistant crashes in libbtrfsutil's
           # UNPRIVILEGED subvolume iteration (btrfs-progs 6.17.1, fixed
           # upstream after 6.17.1) but runs fine as root — which is how the

modules/nixos/oom.nix

@@ -14,9 +14,28 @@
 # the build step, keep the session". nixpkgs default-enables oomd in an
 # inert state (no slices monitored); it's disabled outright below so
 # there is exactly one owner of the OOM story.
+#
+# zram sits one layer earlier: a compressed RAM swap that absorbs
+# day-to-day memory pressure (a browser's idle tabs, a big eval's cold
+# pages) before it ever reaches the earlyoom threshold — more headroom
+# on the same RAM, no disk. High swap priority so the kernel fills zram
+# first; any real disk swapfile stays reserved for the hibernate image
+# (BACKLOG #76's other half — hibernation can't resume from volatile
+# zram, so day-to-day paging must not consume the disk swap). zstd @ 50%
+# RAM are the nixpkgs defaults, kept explicit for reproducibility.
 { lib, ... }:
 
 {
+  zramSwap = {
+    enable = lib.mkDefault true;
+    algorithm = lib.mkDefault "zstd";
+    memoryPercent = lib.mkDefault 50;
+    # Beats any disk swapfile's auto-assigned (negative) priority, so
+    # day-to-day paging lands in zram and the disk swap is left for
+    # hibernation. See the header note and BACKLOG #76.
+    priority = lib.mkDefault 100;
+  };
+
   services.earlyoom = {
     enable = lib.mkDefault true;