Platform updates no longer stall on a source file that could not be read

While a platform update is being prepared, every custom type is rebuilt. Until now, if the platform could not read a type's source files at that moment — busy machines during a changeover, a slow answer from another part of the mesh — it went ahead and built the type from whatever it had managed to load. The result was an error message about your code, for code that was perfectly fine: "the name X does not exist", naming a class that was sitting right there in a file the platform simply had not read.

Those invented errors looked exactly like real ones, so the update safety-check treated them as damage caused by the new version and refused to finish. Updates stalled on healthy content, and the only way out was to switch the safety-check off — the very thing it exists to make unnecessary.

Now the platform never builds a type from an incomplete set of source files. If the files cannot be read, it reports exactly that — "the build state could not be determined", naming the read that failed — the affected type shows a retry message instead of "please correct the code", and the update continues. A genuine error in your code is unchanged: it is still reported as an error, and it still holds the update back, which is what the safety-check is for.

Reconnecting…
The connection to the server was interrupted. Trying to restore it…
Trying again…
The connection could not be restored. Reloading the page…
The server was updated. Reloading the page to pick up the latest version.