What we do is a separate Python process, not in Lambda, that polls status. Status is persisted in Redis. It supports lambdas calling other lambdas so dynamic wo

What we do is a separate Python process, not in Lambda, that polls status. Status is persisted in Redis. It supports lambdas calling other lambdas so dynamic workflows just work and can live for longer than 300s. This has been working fine so far but would probably have issues at 10k+.