@voiced with voiced

In Sanskrit (like in Polish, English, Basque and many other languages) stops in a cluster must either be all voiced stops or all unvoiced stops. Also, voicing usually propagates backwards —

**suhRd- सुहृद् + tama- तमॱ → !**suhRttama- सुहृत्तमॱ by kharica

**anuSTubh- अनुष्टुभ् + chandas- छन्दस् → !**anuSTupchandas- अनुष्टुप्छन्दस् "whose metre is anuSTubh अनुष्टुभ्"

**vidyut- विद्युत् + /bhis → vidyud- विद्युद् + /bhis by jhalAJjazonte → !**vidyudbhis विद्युद्भिस्

bhagavat- भगवत् + gItA गीता → !**bhagavadgItA भगवद्गीता by jhalAJjazonte

This is unlike what happens in Basque, where voicing propagates forward —

ez "no" + da "is" → ezta "it is not"

The exception to this general principle is the rule jhaSastathordh..., that carries both voicing and aspiration FORWARD, but only to t त् th थ् —

budh बुध् + /kta → **buddha- बुद्धॱ

Also, voiced stops won't ever be near the /zar —

**suhRd- सुहृद् + /sup' → !**suhRtsu सुहृत्सु "about friends"

The funny consonants are less racist, and, like the vowels, may live happily near an unvoiced serious —

snAna- स्नानॱ

ratna- रत्नॱ

**khaTvA- खट्वाॱ

**zloka- श्लोकॱ