Renaming a column on a table with millions of rows, while customers in three time zones are using it. The expand-and-contract pattern, the locks that catch people out, and what to do when a migration has to be abandoned halfway.

Twelve screenshots an hour per user becomes terabytes faster than anyone plans for. The arithmetic, what compression actually saves on screen content, and why retention is the only lever that scales.

Rate limiting is easy until the clients are your own apps on thousands of machines. Why a single global limit is wrong, what to key on, and how to handle the thundering herd after an outage.

A tracker that loses an afternoon to a dropped connection gets uninstalled. The local queue, idempotent uploads, clock skew, and the conflict rules that decide what happens when two devices disagree.

How to know somebody stopped working without recording what they typed — the system idle APIs on macOS, Windows and Linux, the permissions each one needs, and why the hard part is the policy, not the code.

The media engine encodes 4K screen capture at a fraction of the CPU cost — and is worse at text than x264. What we measured, where each one wins, and why the default is not obvious.

Sync that is perfect for two minutes and half a second out at forty. Why two clocks drift apart, why a constant frame rate assumption is the usual culprit, and how to keep a long recording aligned.

A recorder that says 60fps and delivers 41. Where frames are actually lost between the display server and the file, how to instrument each stage, and which fixes are worth making.

Screen content is not video content. Why a bitrate that looks perfect on a still page turns to mush when you scroll, how resolution and frame rate change the file, and the settings that actually matter.

Store every timestamp in UTC, compute every boundary in the organisation timezone, and never assume a day is 24 hours. The time zone bugs that only appear near midnight and during a DST change.