worsening of performance in repetitions of the same attack

[EmilioPanti]

Hello,
I have a small lab consisting of 1 Kali machine and 3 metasploitable machines:

• Kali can reach Metasploitable01,
• Metasploitable01 can reach Kali and Metasploitable02,
• Metasploitable02 can reach Metasploitable01 and Metasploitable03,
• Metasploitable03 can reach Metasploitable02,

The goal is to reach the Metasploitable03 machine starting from Kali, where the metasploit RPC server is running. A classic attack therefore includes 5 steps:

1. exploit01 to open a session in Metasploitable01
2. autoroute step to reach Metasploitable02
3. exploit02 to open a session in Metasploitable02
4. autoroute step to reach Metasploitable03
5. exploit03 to open a session in Metasploitable03

If I execute the attack once, everything goes almost always well, managing to reach the metasploitable03 machine.
However, if I repeat the same attack (or another equivalent that uses different exploits but which should be successful), I can hardly ever successfully complete the entire attack. In general, it would appear that exploits are less likely to open a healthy session and the open sessions are very unstable.

Note: between one attack and the next, I kill all open sessions and active jobs.

Can anyone explain to me this increasingly "random" behavior of metasploit in case of repeated attacks? Is there any way to mitigate this behavior?