See
With 2026.4.1 the same, before all was fine.
No error or info in the logs.
Same here. Very noticeable on iPhone 26.4. Barely noticeable on Windows 11 Pro. Almost 1.5 seconds to display icons. I use card_mod and custom button cards; but this was fine pre 2026.4. Thank you.
Looks to be delayed icons so could be releated to Simplify icon loading by piitaya · Pull Request #29953 · home-assistant/frontend · GitHub
Same here. We are talking 10-15 sec to load dashboards on iOS. Pre 2026.4 below one sec. 
I can confirm the slow behavior on iOS and Android as well.
I need to correct my previous post. The HA update coincidentally happened at the same time as a hardware failure on my end.
As it turned out, one of my network switches had died – the very switch my HA server was connected to. It was delivering only around 10 Mbit throughput on all ports.
After replacing the switch, all issues were immediately resolved – including the broken background colors in input fields. Everything is working perfectly now, no issues whatsoever. I’m fully satisfied with update 2026.4.2.
I apologize to anyone who spent time looking into my case. Sorry for the confusion! 
Original Post:
Same here. Yesterday evening I updated from 2026.3.4 to 2026.4.2 and since then everything is loading. Every dashboard, every menu, every view. The worst part are background images on dashboards – you can literally watch them slowly render across the screen, pixel by pixel. Same behavior in every app and every browser. Additionally, I noticed two things that might be related:
1. RAM usage: My HA VM in Proxmox is consuming over 90% of its allocated RAM (6 GB), while HA itself only reports ~1.9 GB usage. Increasing the VM RAM from 4 GB to 6 GB didn’t help – it just filled up to 90% again immediately. This strongly suggests a memory leak introduced with 2026.4.x.
2. Network analysis: When loading a dashboard, browser dev tools show that the server responds quickly (waiting: ~629 ms), but then takes ~14 seconds to transfer only 27 KB – resulting in roughly 2 KB/s throughput. This rules out network as the cause and points to a server-side bottleneck. The behavior is identical whether accessing locally via IP or externally.
And another issue: I use darkmode and in settings Menü every field is now with a withe Background an nealry White font color. In darkmode nothing ist readable. I have to Switch to lightmode if I want to read anything.
FYI, this is absolutely normal in Proxmox. It’s marking the RAM as available for immediate consumption & not actually using it. It’ll release that RAM if another VM actually needs it, so you can safely bump it down back to 4GB.
Just upgraded and my login page took a solid minute and half to load.
I noticed there, and on subsequent dashboard that a large number of .js files are all loading sequentially and very slowly. See network timeline.
What’s very interesting is that if I hit the UI directly, instead of nginx proxy, it loads very fast (less than a second)
But I did not update or change nginx, this started with the 2026.4 upgrade.
My issue was caused by the OS 17.2, not the Core update.
Curious is OP @Piti8576 is using a reverse proxy in front of HA?
The reason I suspected OS changes is that my nginx proxy is throwing error about upstream server disconnecting, and it’s occurring for apps/addons as well. However I have not made any changes or updates NPM and it still routes fine to 2 other servers on my network on different boxes. Anything run on my HA server as docker container is acting up when hit via proxy.
The Fix
I downgraded to 17.0
ha os update --version 17.0
Error logs on 17.2 from nginx reverse proxy
2026/05/01 11:24:28 [warn] 205#205: *62 upstream server temporarily disabled while connecting to upstream, client: 2600:6c64:737f:a4ae:84aa:71be:9b92:42f, server: ha.mydomainname.tld, request: "GET / HTTP/2.0", upstream: "http://172.30.232.1:8123/", host: "ha.mydomainname.tld"
2026/05/01 11:24:28 [error] 205#205: *62 upstream timed out (110: Operation timed out) while connecting to upstream, client: 2600:6c64:737f:a4ae:84aa:71be:9b92:42f, server: ha.mydomainname.tld, request: "GET / HTTP/2.0", upstream: "http://172.30.232.1:8123/", host: "ha.mydomainname.tld"
[01/May/2026:11:24:28 -0400] - 504, 200 200 - GET https ha.mydomainname.tld "/" [Client 2600:6c64:737f:a4ae:84aa:71be:9b92:42f] [Length 2498] [Gzip -] [Sent-to homeassistant.local] "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/147.0.0.0 Safari/537.36" "-"
2026/05/01 11:24:45 [warn] 205#205: *62 upstream server temporarily disabled while connecting to upstream, client: 2600:6c64:737f:a4ae:84aa:71be:9b92:42f, server: ha.mydomainname.tld, request: "GET /frontend_latest/70762.16533dcba1df9834.js.map HTTP/2.0", upstream: "http://172.30.232.1:8123/frontend_latest/70762.16533dcba1df9834.js.map", host: "ha.mydomainname.tld"
... many more 504s...
2026/05/01 11:24:45 [warn] 205#205: *62 upstream server temporarily disabled while connecting to upstream, client: 2600:6c64:737f:a4ae:84aa:71be:9b92:42f, server: ha.mydomainname.tld, request: "GET /frontend_latest/26767.e1ef871fec1c0b33.js.map HTTP/2.0", upstream: "http://172.30.232.1:8123/frontend_latest/26767.e1ef871fec1c0b33.js.map", host: "ha.mydomainname.tld"
[01/May/2026:11:24:45 -0400] - 504, 200 200 - GET https ha.mydomainname.tld "/frontend_latest/51308.2bfcd39933750357.js.map" [Client 2600:6c64:737f:a4ae:84aa:71be:9b92:42f] [Length 10535] [Gzip -] [Sent-to homeassistant.local] "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/147.0.0.0 Safari/537.36" "-"
... a couple successful calls...
[01/May/2026:11:24:48 -0400] - 201 201 - POST https ha.mydomainname.tld "/api/webhook/139c79007fffc847333e0d62487508547cf729308a3c672321afdcc35f6d674f" [Client 2600:6c64:737f:a4ae:f3da:52c1:b6cb:f68c] [Length 16] [Gzip -] [Sent-to homeassistant.local] "Home Assistant/2026.4.4-21576 (Android 16; Pixel 10)" "-"
2026/05/01 11:24:50 [warn] 205#205: *62 upstream server temporarily disabled while connecting to upstream, client: 2600:6c64:737f:a4ae:84aa:71be:9b92:42f, server: ha.mydomainname.tld, request: "GET /frontend_latest/85545.847aa1acac65a4f5.js HTTP/2.0", upstream: "http://172.30.232.1:8123/frontend_latest/85545.847aa1acac65a4f5.js", host: "ha.mydomainname.tld"
2026/05/01 11:24:50 [error] 205#205: *62 upstream timed out (110: Operation timed out) while connecting to upstream, client: 2600:6c64:737f:a4ae:84aa:71be:9b92:42f, server: ha.mydomainname.tld, request: "GET /frontend_latest/85545.847aa1acac65a4f5.js HTTP/2.0", upstream: "http://172.30.232.1:8123/frontend_latest/85545.847aa1acac65a4f5.js", host: "ha.mydomainname.tld"
[01/May/2026:11:24:50 -0400] - 504, 200 200 - GET https ha.mydomainname.tld "/frontend_latest/85545.847aa1acac65a4f5.js" [Client 2600:6c64:737f:a4ae:84aa:71be:9b92:42f] [Length 4896] [Gzip -] [Sent-to homeassistant.local] "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/147.0.0.0 Safari/537.36" "-"
pattern repeats, and identical for other containers running as addons.
Has anyone found a solution to this that isn’t downgrading? The UI is significantly slower on desktop and takes several minutes to load on mobile, after the upgrade to 2026.4. Nothing else has changed.
I can reproduce the slowdown and wanted to share what helped in my case.
My setup: HAOS 17.3, Core 2026.5.0.
UI was laggy, and homeassistant.local was clearly slower than direct IP access.
What I found:
- A failing integration (in my case Immich setup with 'http://immich.local' instead of ip-address) was producing repeated DNS timeout errors.
- I also had reverse-proxy forwarding warnings in the system log.
- Both added noise/load and made troubleshooting harder.
What fixed it for me:
- Reinstalled the failing Immich integration (stopped repeated DNS errors).
- Added proper reverse-proxy settings in configuration.yaml
http:
use_x_forwarded_for: true
trusted_proxies:
- 192.168.178.9 --> my nginx proxy
- Restarted Core.
After that, CPU dropped from ~50% to ~20–25%, UI responsiveness improved a lot, and those log errors disappeared.
If you’re seeing similar symptoms, also test HA via direct IP (instead of homeassistant.local) to rule out mDNS-related latency.
I'd be curious if you at least confirm this helps or not, it's easy to move the OS forward and back.
If does fix it I'll open an issue with OS.