### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

Since I've updated my litellm-proxy installation to v1.89.0, it's logging these two errors on practically every LLM call:

``` parallel_request_limiter_v3.py:2415 - TTL preservation failed, falling back to regular pipeline: unsupported operand type(s) for +: 'float' and 'str'

redis_cache.py:632 - LiteLLM Redis Caching: async set() - Got exception from REDIS unsupported operand type(s) for +: 'float' and 'str' Writing value={'timestamp': 1781617633.1508543, 'response': '{...}'} ```

This also constantly trips the Redis circuit breaker, meaning that caching is effectively disabled on my proxy.

This is on a litellm-proxy deployment on kubernetes, using the `berriai/litellm` chart version 0.1.1. , using 3 pods/replicas.

Redis caching is enabled. The actual Redis I'm using is an `Azure Cache for Redis` instance on Azure, using Redis 6 and with SSL enabled and set to mandatory. Caching was working correctly with litellm-proxy image `main-v1.83.14-stable.patch.3` which was what I had before upgrading.

### Steps to Reproduce

…