Twitter spam bots that offer to sell me bitcoin or start following me with a beautiful profile picture who just opened their twitter account and I don't know are pretty obviously bots to me. The former could be found using the firehose, but the latter wouldn't be detected until they tweeted something in the firehose.
If the firehose included DMs, new accounts (with all available signup and verification data), etc. then the firehose would be useful. Detecting bots using the firehose would only detect bots based on suspicious tweet content and nothing else.
Also, the firehose is a 7Mbps (compressed) stream. It takes substantial compute power to run just sentiment analysis and substantial storage to keep heuristics on that volume of data. Musk would need to spend some time and money and hire some people to just make sense of the full firehose.
Source: I worked for GNIP in 2012. GNIP used to re-sell the twitter firehose and was a twitter partner before they were acquired by twitter in 2014.
What data would that be though ?
1) Email addresses would be valid since they are used during the verification process.
2) Location data is only available on mobile which bots obviously won't be using.
3) IP address data would be filtered through compromised hosts.
4) Interaction data is probably the key one. But this is trivially forged.
The type of bots that are most common on Twitter are those from well funded interests e.g. governments, companies or technical ones e.g. hackers.
People who know what they are doing and would be building something largely imperceptible from ordinary users.