Currently flox activate -m run -s in shell 1 then flox activate -m dev -s in shell 2 then exiting shell 1 will tear down shell 2's services because they both use separate FLOX_ENV (thus also separate activations.json and flox-watchdog instances) but the same _FLOX_SERVICES_SOCKET paths.

Possible solutions:

• Remove the mode specific component of the path from what we're hashing to determine activations.json path, and then modes will not tear down services for other modes
• Error if an environment is activated in multiple modes

Original thread: https://flox-dev.slack.com/archives/C05P6A5J6U8/p1731431644112099