cwnd differences between ssh tunneled application versions (one has pausing in comms)

I am comparing two different versions of the same application that are running on the same server and client and I am trying to understand the results of ‘ss --info --tcp ‘.

One of them has cwnd always around 10 or so. The other cwnd is as high as 40 or 50.

The higher one suffers from pauses in the communication channel.

Would the difference have to do with an application setting or perhaps with how much data is being transferred? Both transfer lots of information but perhaps one sends lots of smaller packets vs the other with larger packets. I’m trying to help track down the cause of the pauses. Thx for the help!

Here’s the full output for each (and cwnd for the higher one is 36 in this output):

Version that works well: ESTAB 0 44 Xxx:ssh Yyy:aaa
cubic wscale:6,8 rto:339 rtt:138.692/0.298 ato:40 mss:1228 rcvmss:1228 advmss:1448 cwnd:10 ssthresh:18 bytes_acked:4138781 bytes_received:257537 segs_out:9259 segs_in:8761 send 708.3Kbps lastsnd:3 lastrcv:24 lastack:131 pacing_rate 1.4Mbps unacked:1 retrans:0/32 reordering:4 rcv_rtt:141 rcv_space:28960

Version with pauses: ESTAB 60 88 Xxx:ssh Yyy:bbb
cubic wscale:6,8 rto:338 rtt:137.149/0.085 ato:70 mss:1228 rcvmss:1228 advmss:1448 cwnd:36 ssthresh:36 bytes_acked:10281665 bytes_received:611629 segs_out:21355 segs_in:26563 send 2.6Mbps lastsnd:72 lastrcv:2 lastack:2 pacing_rate 5.2Mbps unacked:2 retrans:0/154 rcv_rtt:138 rcv_space:28960