Bluetooth off at every boot, slow to reconnect, and no headset battery level
The main.conf content is unusually accurate. I checked every key against bluez master src/main.conf and each one is in the section the record puts it in. [General] holds Class (0x000100), DiscoverableTimeout, ControllerMode (default dual), FastConnectable (default false), Experimental (default false), KernelExperimental (default false). [Policy] holds ReconnectAttempts (7), ReconnectIntervals (1,2,4,8,16,32,64), AutoEnable (default true), ResumeDelay (2). The Arch Bluetooth page confirms the AutoEnable history exactly ("As of bluez 5.65, BlueZ' default behavior is to power on all Bluetooth adapters when starting the service or resuming from suspend"), the ControllerMode=bredr fallback for BR/EDR-only devices, the DiscoverableTimeout=0 recipe, and that battery reporting for headsets rides on the D-Bus experimental interfaces ("enabling D-Bus experimental interfaces currently allows to report battery level for old headsets"). The --experimental drop-in and /usr/lib/bluetooth/bluetoothd path are right, as is the rfkill escalation. ONE defect, in the Omarchy paragraph: omarchy-restart-bluetooth does exist and *is* reachable at Update > Hardware > Bluetooth in bin/omarchy-menu, but its entire body is rfkill unblock bluetooth; rfkill list bluetooth. It never touches bluetooth.service. So the record's advice to use it "after changing this file rather than fighting the daemon by hand" is backwards: it will not reload main.conf, and the user will conclude their edit did nothing. (Minor: the text refers to "the danger note" about the pacman-owned file, but no such note is in the record, and pacman lists main.conf in bluez's backup array, so edits survive upgrades and land as .pacnew, which is worth stating instead.)
/etc/bluetooth/main.conf is owned by the bluez package, so editing it means every bluez upgrade drops a main.conf.pacnew next to it and your settings silently stop matching upstream defaults. Run pacdiff (from pacman-contrib) after upgrades and merge. Experimental and KernelExperimental deliberately turn on code paths BlueZ has not declared stable. If pairing or audio becomes flaky after enabling them, they are the first thing to revert. FastConnectable = true increases idle power consumption, which matters on a laptop. Setting ControllerMode = bredr disables Bluetooth LE entirely, which will break LE-only peripherals (many modern mice, trackers and keyboards).
Symptom
Three complaints that all trace back to the same file. (1) Bluetooth is dead after every boot: bluetoothctl show prints Powered: no and you have to run power on by hand or toggle it in the bar every single time. (2) Your headset or mouse takes 10-30 seconds to reconnect when you turn it on, where Windows reconnects near-instantly. (3) The headphones connect fine but their battery level never appears anywhere: upower -d shows no percentage and the bar shows nothing.
Cause
/etc/bluetooth/main.conf governs adapter power-on, reconnect behaviour and which BlueZ features are exposed, and its shipped defaults are conservative. [Policy] AutoEnable defaults to true since bluez 5.65, but a stale AutoEnable=false left over from an old guide (or a persistent rfkill soft block) keeps the adapter down. [General] FastConnectable defaults to false, which leaves the page-scan interval long and makes peripheral-initiated reconnection slow. Battery reporting for most headsets is served over BlueZ's D-Bus *experimental* interfaces, which are off unless Experimental = true is set, so upower and every bar widget that reads it has nothing to show. Separately, [General] ControllerMode defaults to dual, and BR/EDR-only devices sometimes fail to associate until it is forced to bredr.
Fix
Replace the Omarchy paragraph with:
**On Omarchy**, still apply the change with sudo systemctl restart bluetooth.service. The menu item *Update > Hardware > Bluetooth* runs omarchy-restart-bluetooth, which despite the name only does rfkill unblock bluetooth and prints rfkill list bluetooth. It does not restart bluetoothd and will not pick up a changed main.conf. Use it for the rfkill-soft-block case (the escalation later in this record), not for config changes:
sudo systemctl restart bluetooth.service
bluetoothctl show # confirm Powered: yes and the new settings took
Also replace the "see the danger note" aside with the accurate statement: /etc/bluetooth/main.conf is owned by the bluez package but is listed in its backup array, so your edits are preserved across upgrades and a changed upstream default arrives as /etc/bluetooth/main.conf.pacnew. Merge it with pacdiff after a bluez update rather than assuming your file is current.
Verify
bluetoothctl show reports Powered: yes immediately after a cold boot with no manual intervention. busctl --system introspect org.bluez /org/bluez/hci0/dev_XX_XX_XX_XX_XX_XX | grep -i battery shows the Battery1 interface once Experimental is on. upower -i on the headset device prints a percentage: line. Time a reconnect before and after FastConnectable = true.