Are you sure that failures are because of RUM injection? RUM is catching all exceptions in the console log, because of that DT can show you whole exception messages. Such exceptions are thrown than by dtjsagent again so sometimes it looks like the agent is the source of errors true stack trace is from the moment when the other JS file starts. It is possible that even without JS agent you had those errors but you didn't know. If this is intermittent and not constant, it looks like what I said.
SignalR should be transparent for agents, you will see the loading of this file and errors but nothing more as far as I know.
After disabling the RUM we haven’t seen these 500 errors but when we enabled it back they occur intermittently.
We have single tenant for multiple environments including both prod and test environments. Is it possible to exclude the RUM monitoring for the signalR only for a specific process groups rather than global?