Elastic frequently disconnects when using a large number of widgets

[godmar]

Describe the bug
We are using a single elastic layout for tuning all our subsystems. When activating it with the robot it is unusable because it disconnects, reconnects, disconnects, reconnects over and over.

When we remove all but one tab, and thus reduce the number of topics Elastic subscribes to, it becomes usable again. (We haven't found the actual point when it becomes usable again - we know it's usable with one tab and 4 widgets and unusable with all.)

Expected behavior
It should not disconnect.

Desktop (please complete the following information):

• OS: Windows
• 2026.1.2

Additional context

I asked Codex 5.3 to look at this codebase. Here is its analysis. I haven't verified any of its claims, although they seem consistent with the behavior we are seeing. Codex proposed this fix which I also haven't vetted.

I'm leaving this bug report here mainly to see if others have experienced similar issues. Our hope was to use Elastic as our main dashboard during competition, but we can't if it's not stable.

[Gold872]

Thanks for reporting this issue. It sounds like this issue is caused by performance issues on your laptop due to the number of subscriptions there are. I implemented a few of the suggested fixes in this PR, can you try out the build from that PR and see if it resolves the issues you're seeing? I want to avoid increasing the timeout, since 1 second should be enough

[godmar]

Thanks for reporting this issue. It sounds like this issue is caused by performance issues on your laptop due to the number of subscriptions there are. I implemented a few of the suggested fixes in this PR, can you try out the build from that PR and see if it resolves the issues you're seeing? I want to avoid increasing the timeout, since 1 second should be enough

Thank you. The laptop is an average 2024ish model. Keep in mind it has to run driver station, advantagescope and various other software. I'm not sure if the student working on this is set up to run elastic from a PR easily - is there a way to download a prerelease build for testing?

[Gold872]

Yes, you can download one here: https://github.com/Gold872/elastic_dashboard/actions/runs/22630863045

[Gold872]

Have you been able to test whether the development build fixed the issue?

[godmar]

Have you been able to test whether the development build fixed the issue?

Yes. We did not see any improvement, it keeps disconnecting, also on more powerful laptop hardware.

[Gold872]

This sounds like it's likely a networking related issue, I could try increasing the timeout, but that feels like a fix that's evading the actual problem. I'm going to ping @PeterJohnson and see if he has any ideas

[godmar]

This sounds like it's likely a networking related issue, I could try increasing the timeout, but that feels like a fix that's evading the actual problem. I'm going to ping @PeterJohnson and see if he has any ideas

It could be networking related; I need to point out that AdvantageScope running on the same machine (and presumably also subscribing to the same network table server) does not have any problems.