What changed in each version.

The app checks for updates once a day and asks before downloading. You can also check now with Check for updates… in the tray menu.

0.19.0

October 10, 2026

With “Suggest the app in the foreground” on in Settings, the shortcut saves with the lag only the file name of the program in the foreground, such as the game's, never its folder, window title or connections. In the details of a marked lag, the name shows as a suggestion you can correct or clear, and the report does not treat it as the cause.
In Settings, each Alt key leads to one control on each tab, and the pt-BR label for recording lag spikes and packet loss is no longer cut off with large text.
In the details of a marked lag, you can give only when the problem started, without the end: when you marked it, 30 s before, 1 min before, or a typed time. The report shows this approximate start apart from the marker, says the end was not given, and analyzes the measurements around that start.
The report shows whether Windows counted errors or discarded packets on each network adapter before, during and after the lag. Counts that could not be read appear as unavailable, not as zero.
DNS lookups also ask 8.8.8.8 directly, besides 1.1.1.1 and the system resolver. The report and the dashboard's DNS chart show each server separately, and a server that does not answer does not get in the way of measuring the others.
When the system resolver and every direct server fail together, the report says so and whether the external pings answered at that moment, without claiming that all DNS was down. At the top of the dashboard, DNS counts only the system resolver's failures; the direct servers' failures appear in the chart.
If you change DNS or turn on a VPN after the lag, the report keeps the network configurations observed during the incident and their timestamps, including when rebuilt. Failed reads appear as gaps, and the copy for sharing numbers the addresses and interfaces in this history.
Opening the app again while it is closing says that Something Just Lagged is already running, instead of quitting with no dashboard and no notice. While the app is open, a new start still opens the dashboard.
Check automatically once a day also counts the time the PC spent asleep or hibernating: a laptop that slept all night looks for updates within an hour of waking, without waiting for 24 hours of being on.
The report shows when the Cloudflare datacenter answering the HTTPS check changes, with the time of each change. The app keeps only the datacenter code and discards the rest of the response, including your public IP address.
When only the system resolver fails, the report's "Reading near the incident" section no longer names the PC's DNS servers. That list is read when the report is generated and may not be the one in use during the lag, for example if you turned on a VPN afterwards; it is still in the observations, with that caveat.
When the app cannot save the reading in which a network adapter went down or came back, because the disk is full or an antivirus locked the file, it saves that reading again at the next check, and the report counts the interface status change.
When saving a measurement is cut off midway, for example by a full disk, the next measurement is no longer lost along with the incomplete one.
The app starts normally when a section or list in a hand-edited config.json is null: it goes back to the default. An empty list still turns that measurement off.
A dashboard port outside 1024 to 65535 in config.json now uses 8765, instead of keeping the app from starting or leaving the dashboard off.
The update window says what to do when an update stops for a reason you can fix, such as closing the incident window before installing, instead of asking you to check your connection. If the offered version changed or was withdrawn, the Download and Install and restart buttons go away, and the window asks you to look for the latest version with Check for updates.
A downloaded update that got corrupted on disk no longer makes every Download or Install and restart attempt fail until someone deletes the file by hand. After the first failure, Download fetches the update again.

0.18.0

October 3, 2026

Pressing the shortcut again or choosing Mark incident in the tray menu within 2 seconds of marking a lag does not create another incident or notification, and keeps the first notification’s destination. Once 2 seconds have elapsed since the mark, you can mark another lag; ignored presses do not extend the wait.
After the free trial, the tray menu and the dashboard history show how many recorded incidents await a license, counting both marked and automatic incidents, without new notices or windows. The count offers a purchase link and disappears when you activate the license, while the app generates the reports.
Moving the pointer over dashboard charts to check values is smoother, even with All measurements open. Measurements stay the same.
In All measurements, a chart's screen-reader description no longer says there are no measurements after they arrive.
In the 24-hour and 7-day ranges, the dashboard fetches data less often to use less CPU: charts, hop losses, Twitch cards and markers may take up to 2 minutes to update. Shorter ranges still refresh every 20 seconds, recording and extension status keep their previous frequency, and changing the range, returning to the tab or following a capture does not wait for that interval.
In the 1-hour to 7-day ranges, dashboard refreshes with Twitch measurements are faster and use less CPU: after the first read, the app reads only the new measurements instead of reading every day in the range again. Results stay the same.
The dashboard uses less memory and CPU when analyzing long periods with Twitch measurements. Results stay the same.
The dashboard reads measurement files fewer times to update its counters and general information, keeping that refresh at every 20 seconds.
New Twitch observations take less disk space: the app no longer writes fields with no value, keeping the same information in the dashboard and reports.
The app uses less temporary memory to send data to the dashboard, keeping the same measurements.
When the extension sends several Twitch observations together, the app writes them in a batch, with fewer disk writes and the same measurement order. Pausing recording still waits for observations already accepted.
When the app is unavailable, the extension asks you to open Something Just Lagged and check the connection, without naming an old version. The new notice requires updating the browser extension.
The guide bundled with the app points Firefox users to Mozilla Add-ons and keeps the manual steps for Chrome and Edge. ZIP names use the extension version, and only manually loading the Firefox package is described as temporary.
Installer and update downloads report the file size from the start. This lets the browser show the installer download progress.

0.17.1

October 3, 2026

In the incident report, a response from a website ends the interval without external replies even if the transfer fails afterward. The report now agrees with outage detection and history in these cases.
In the dashboard, under Twitch in your browser, inside Connect the browser extension, the Disconnect extension button makes the current pairing code stop working: the extension stops sending observations until you copy a new code and paste it there.
When a network connection on the PC disconnects or comes back, the app records the change on the next reading instead of waiting for the routine record every 60 seconds. The report also counts these changes observed between routine records.
With the dashboard open on 6 h, 24 h or 7 days, the refresh every 20 seconds no longer reads every measurement of the period again for the latency, DNS, HTTPS and traffic charts: the app keeps the points it already computed for each day and reads only what was measured since the last refresh. On 1 h, it reads only the last hour. The first time a period opens takes a little longer; the next refreshes are much faster and the charts stay the same. The Twitch analysis still reads its measurements again on every refresh on 24 h and 7 days.
On 10 min, the first time the dashboard opens after the app starts is faster, and on 7 days every refresh uses less CPU: the app reads only the time of each stored measurement and reads in full only those it needs. The charts stay the same.
With the app closed, the browser extension stops trying to reach it every 4 seconds and lets the browser rest. It also no longer wakes up every 30 seconds when there is nothing to send.
The browser extension does less work while Twitch is open, even with several players at once.
While the app refuses the pairing code, the extension does not pile up observations in the browser: those older than 12 seconds are dropped and counted as lost, as already happens with the app closed.
When the app closes or shows an unexpected error window, app.log in the data folder (tray menu, Open data folder) keeps the full error, so you can send it along when reporting the problem.
app.log no longer grows without limit when the app stays open for days: past 2 MB it becomes app.log.old and starts over, so the two together take at most about 4 MB.
The Wi-Fi measurement is only recorded when something changes (signal, network, access point, channel or state), once a minute with no change, and on every reading for the 2 minutes after you mark a lag or the app detects one: on a PC on a wired connection with Wi-Fi present but unused, a day goes from about 17,000 identical lines in the data folder to about 1,440, and with Wi-Fi in use the signal moves almost all the time, so the measurement stays every 5 seconds. In the report, the Wi-Fi sample count is lower over stretches where nothing changed, and the signal median now takes into account how long each reading lasted, not only how many there were.
The 7-day free trial counts by the latest time the app has seen on this PC: setting the computer's clock back does not reopen or extend the trial. If the clock is set before the day the trial started, incident reports wait until you set the date and time and open the app again; recording continues.
If the app could not identify the computer on first launch and can later, it asks for the activation code again, or offers the free trial, instead of keeping the provisional identification.
One internet connection can start up to 3 free trials every 14 days. Asking for the trial again on the same PC does not count: it returns the period that was already running.
Starting the free trial more than once at the same time on the same PC uses only one of the connection’s 3 slots every 14 days.
When automatic update checks fail for over a week, the app tells you once from the tray. Clicking the notice opens Check for updates, which suggests trying again and, if it keeps failing, downloading the latest version from the website without losing your data or license.
PC traffic no longer double-counts what passes through a VPN or a virtual adapter (Hyper-V, VMware, WSL): the totals count physical adapters only, and the per-adapter detail is still kept.
With cable and Wi-Fi connected at the same time, the app measures the router of the connection Windows is actually using. With a VPN on, it keeps measuring your network's router, not the VPN server.
When the server that confirms a site's certificate cannot be reached, HTTPS measurements no longer record that as a failure of the site. Revoked certificates are still rejected.
When there are no router measurements during the incident window, the report states that absence without adding a warning about measurements with no recorded destination.
The report no longer says the analysis focus overlaps the PC's sleep or reconnection when the symptom starts and ends at the same time.
The note in the PDF report preserves #, backticks and **, as in "ticket #4521". This fix does not extend to recorded values inside the report's Markdown blocks, such as Wi-Fi names in the body or tables, which can still lose these characters.
If the download server fails to retrieve the installer, the website shows the download unavailable page instead of an unexpected error.

0.16.0

October 2, 2026

In the dashboard's report window, Download for sharing makes a copy of the ZIP without your Windows user name, the computer name, the Wi-Fi network names or the access point addresses (BSSID), to send to your provider, forums or an AI. The report in that copy opens by saying it is a copy for sharing.
In the tray menu, Save copy for sharing… saves that same copy of the last incident to the folder you choose. Open last incident bundle… still shows the full ZIP, with those details.
In the dashboard, Download report (PDF) downloads just the incident's report.pdf, the sheet to attach to a support ticket with your provider or a complaint to the regulator, without opening the ZIP. The button appears on the selected incident and in the report window, which now lists report.pdf among the files in the ZIP.
When an incident doesn't fit on one page, report.pdf continues on a second page with the suggested next test and the sampling coverage in the analysis focus; anything beyond two pages stays in report.md. The first page says the report continues on the second, and the second repeats the incident type and the preserved window, so it still identifies the incident if printed on its own.
In the dashboard, the Latency over time chart now has the same soft gradient under each line as the charts with one or two lines, even when it shows the router and several destinations at once.
In the copy for sharing, DNS servers on your own network, such as the router, show as dns-1, dns-2, because the router's IPv6 address identifies your home. Your provider's DNS stays in the copy when the app can tell it apart from the router; when it can't, as in incidents recorded before this version, it is numbered too.
In the copy for sharing, when the path trace goes over IPv6, hops on your own network show as router-1, router-2, or with the same dns-1 when the hop is your DNS server. If the first of the External targets is an address on your own network, it is replaced too, in the trace and in the pings. Your provider's and the internet's hops stay when the app can tell them apart from your network; when it can't, as in incidents recorded before this version, every IPv6 hop except the destination is numbered.
On first launch, the dashboard welcome draws what the shortcut keeps: the 5 minutes before you press it and the 2 after. It also shows the tray icon to look for.

0.15.0

September 26, 2026

Something Just Lagged now has its own icon in the Start menu, the taskbar, File Explorer, the installer and the app's windows, the same one as the website and the browser extension.
When you change the details of a marked lag after its report is ready, the rebuilt report stays in the language it was first written in, even if the app language changed in Settings.
An outage that crosses midnight is judged by the same Minimum duration and destinations on both sides of midnight. If the Minimum duration changes in Settings the next day, the 30-day history still shows one outage instead of a stray part before midnight, below both minimums.
In the history PDF, the Router during the outage column counts the same probes the automatic incident uses to tell an internet outage from a local network outage, so the two no longer disagree.

Back to the website