We are seeing issues where fonts flood as a result of various things that are either done by an application or by Cypress.
Chrome has some behavior that triggers reloading of in-memory font assets when global style sheets are updated. This has the effect of triggering all font requests to reload. Even though they are in-memory, this produces a flood of CDP events which backs up proxy correlations and bloats the test replay data being captured.
We can fix this by ensuring that the in memory cache is only used in a given context via the ScopeMemoryCachePerContext feature. This will ensure that the resource is in a local cache and it won't replay the CDP events (https://github.com/chromium/chromium/blob/118.0.5993.117/third_party/blink/renderer/platform/loader/fetch/resource_fetcher.cc#L2650-L2652).
Note that we initially tried to fix this issue via #27860 when we thought it was something that Cypress was inherently doing as part of actionability. This did not resolve the issue as frameworks in the apps themselves can be causing the problem as well. As a result as a part of fixing this, we should undo: #27860
We are seeing issues where fonts flood as a result of various things that are either done by an application or by Cypress.
Chrome has some behavior that triggers reloading of in-memory font assets when global style sheets are updated. This has the effect of triggering all font requests to reload. Even though they are in-memory, this produces a flood of CDP events which backs up proxy correlations and bloats the test replay data being captured.
We can fix this by ensuring that the in memory cache is only used in a given context via the
ScopeMemoryCachePerContextfeature. This will ensure that the resource is in a local cache and it won't replay the CDP events (https://github.com/chromium/chromium/blob/118.0.5993.117/third_party/blink/renderer/platform/loader/fetch/resource_fetcher.cc#L2650-L2652).Note that we initially tried to fix this issue via #27860 when we thought it was something that Cypress was inherently doing as part of actionability. This did not resolve the issue as frameworks in the apps themselves can be causing the problem as well. As a result as a part of fixing this, we should undo: #27860